“Crop Recommendation System using Soil and

Weather Content”

A major project report submitted in partial fulfillment of the

requirement for the award of degree of

Bachelor of Technology

in
Computer Science & Engineering / Information Technology

Submitted by
Kshitij Vats(201361) Arpit Aggarwal(201474)
Under the guidance & supervision of
[Link] Dhiman

Department of Computer Science & Engineering and

Information Technology
Jaypee University of Information Technology, Waknaghat,
Solan - 173234 (India)
 CERTIFICATE

This is to certify that the work which is being presented in the project report titled “Crop
Recommendation system using soil and weather content” in partial fulfilment of the
requirements for the award of the degree of [Link] in Computer Science & Engineering and
Information Technology and submitted to the Department of Computer Science & Engineering,
Jaypee University of Information Technology, Waknaghat is an authentic record of work carried
out by “Kshitij Vats(201361)” and “Arpit Aggarwal(201474) ” during the period from from
August 2023 to May 2024 under the supervision of [Link] Dhiman, Department of Computer
Science and Engineering, Jaypee University of Information Technology, Waknaghat.

Kshitij Vats(201361) Arpit Aggarwal(201474)

The above statement made is correct to the best of my knowledge.

Supervisor Name: [Link] Dhiman

Designation: Associate Professor

Department: Computer Science & Engineering and Information Technology

Jaypee University of Information Technology, Waknaghat

(I)
 Candidate’s Declaration

I hereby declare that the work presented in this report entitled Crop Recommendation system
using soil and weather content” in partial fulfillment of the requirements for the award of the
degree of Bachelor of Technology in Computer Science & Engineering / Information
Technology submitted in the Department of Computer Science & Engineering and Information
Technology, Jaypee University of Information Technology, Waknaghat is an authentic record
of my own work carried out over a period from from August 2023 to May 2024 under the
supervision of [Link] Dhiman Department of Computer Science and Engineering, Jaypee
University of Information Technology, Waknaghat.

The matter embodied in the report has not been submitted for the award of any other degree or
diploma.

Student Name: Kshitij Vats Student Name: Arpit Aggarwal

Roll No.: 201361 Roll No.: 201474

This is to certify that the above statement made by the candidate is true to the best of my
knowledge.

(Supervisor Signature with Date) Supervisor Name: [Link] Dhiman

Designation: Associate Prof.
Department: CSE
Dated:

(II)
 ACKNOWLEDGEMENT

I want to thank God for always giving me the necessary blessings and directives which have
lighted up my schooling career.

Special thank you goes to my supervisor, Assoc. Prof. Dr. Pankaj Dhiman, of CSE Department,
Jaypee University of Information Technology, [Link] contribution of Dr. Dhiman to
the success of this task cannot be underestimated. His intelligent comments, critique, and deep-
rooted knowledge on machine learning were precious for me. He maintained his composure,
offered intellectual guidance and continued moral support for me every day as I maneuvered
through this maze of a research project.I would like to thank Dr. Dhiman for spending hundreds
of hours on reading several versions, providing helpful advice and insisting that the work be of
best quality possible.

I also express my gratitude for my parents, colleagues, and educators because their tolerance,
trust, and assistance during this study process have never been wavered. The whole success story
can be attributed to their belief in me.

Kshitij Vats (201361) Arpit Aggarwal (201474)

(III)
 TABLE OF CONTENT

TOPIC PAGE NO.

Certificate...................................................................................................I

Candidate’s Declaration............................................................................II

Acknowledgement....................................................................................III

Table of Content.......................................................................................IV

List of Figures..........................................................................................VI

List of Abbreviations...............................................................................VII

Abstract...................................................................................................VIII

Chapter 1: Project Introduction.............................................................1

Project Introdction 1.1................................................................................1

Problem Statement 1.2…...........................................................................10

Objectives 1.3............................................................................................12

Signifance and Motivation 1.4..................................................................14

Organization 1.5........................................................................................15

Chapter 2:Literature Survey..................................................................17

Overview of Relevant Literature 2.1….....................................................23

Key Gaps in Literature 2.2.........................................................................25

Chapter 3:System Development….........................................................29

Requirements and Analysis 3.1…............................................................. 29

Project Design and Architecture 3.2…………………………………….. 29

Data Preparation 3.3……………………………………………………....34

Implementation 3.4 ............................................................................................. 35

Key Challenges 3.5............................................................................................. 47

Chapter 4:Testing… ........................................................................................49

Testing Strategy 4.1............................................................................................ 49

Test Case and Outcomes 4.2.......................................................................51

Chapter 5:Results and evaluation…........................................................54

Results 5.1… ....................................................................................................... 54

Chapter 6: Conclusions and Future Scope ................................................. 63

Conclusion 6.1 .................................................................................................... 63

Future Scope 6.2 ................................................................................................. 64

References ..................................................................................................66

(V)
LIST OF FIGURES

(VI)
LIST OF ABBREVATIONS

(VII)
 ABSTRACT

This project attempts to make a predictive crop selection scheme using artificial neural networks
utilizing weather and soil information in order to improve farm yields.

It is worth noting that machine learning algorithms are used for doing predictive modelling with
a combination of in-depth studies on the composition of the soil, historical and current weather
conditions. Various agricultural databases are employed in these models to establish the best
crops fitting to different environments. Some of them are regression and classification
algorithms.

The system helps farmer to easily make a decision by adapting with varying soil and weather
conditions. It provides customised guidelines which are aimed at promoting sustainable practices.

In turn, the project contributes towards crop yield, resource utilization, and farm management in
general. Reliability and efficiency of the system in optimizing choice of crop through interaction
between weather and soil content is ensured by rigorous testing and comparing it against other
schemes.

Ultimately this Crop Recommendation System employs state of the art machine Learning
technology in order to create a flexible and responsive system for making optimum choice of
crop dependent upon weather and contents of soils and as such constitute a useful element
towards modernization

(VIII)
Chapter-1 PROJECT INTRODUCTION

1.1 PROJECT INTRODUCTION

It is important that crop recommendation should be carried out properly that it will
improve agricultural productivity towards sustainability in agricultural process.
This research project recommends a Crop Recommendation system using machine
learning algorithms to give specific crop recommendations based on soil and
climatic information.

This system seeks to help the farmers and agricultural specialists to make
intelligent choices on crop selections by using the potential of data analysis and
predictive modeling. Choosing the right crops to be grown is difficult as it entails
many aspects like the composition of the soil, weather condition among others.
The traditional crop recommendation methods usually depend on expert’s
knowledge or experience which is prone to personal bias and laborious.

Due to advances in Machine learning, and data analytics, the possibility arise of
designing a smart system which will automates and improve this undertaking. For
the purpose of this study, a Crop Recommendation system employing different
machine learning algorithms on soil and weather parameters will be developed.
The system takes into account factors like the quality of soils, nutrients content in
the land, temperature, level of rainfall and humidity, among others; and determines
the best crops to grow in a given region. In order to obtain these results, a database
containing information on soil composition, previous weather data as well as
corresponding crop yields shall be built up and prepared for analysis purposes.

Dataset will be used for training the machine learning model, either a regression
or a case of clustering depending on the input variables and their connection to
crop suitability. This will help in ensuring that the recommendations are reliable

1
and accurate, and also provide the models with necessary performance metrics.
The system will be presented in the form of an easy-to-use interface that accepts
data either on user’s location or based on soil and weather.

characteristics. After the system has analyzed the inputs, it will come up with
personalized recommendations for growing crops in the particular region.

The suggestions shall appear in simplified language, which will enable farmers to
choose suitable [Link] crop recommendation system can transform the
agriculture sector by using Machine learning capabilities and the power of data
analysis. It will also educate farmers, provide necessary information for crop
choice, help create a sustainable form of agriculture. Prolonged testing and
verification involving rigorous experiments will be carried out for the purpose of
ascertaining that the system is both reliable and economical. The aim of this
project is to come up with a Crop Recommendation system that will use the latest
Machine Learning technologies so as to give precise and specific crop advices.

Our Proposed prediction system aims to help farmers as follows-

•Optimized Crop Selection:

Our envisaged model would have transformed choice of crops for individual
farmers on a personal basis. The system utilizes actual weather information and
soil makeup in real time and gives farmers individual advice suitable for getting
high yields. Thus this selective approach to choosing crops is meant to increase
both farm output and earnings in a crop- growing field.

•Resource Efficiency:
This system is more than just recommending since it considers the distinct natural
circumstances in every farm. It enhances resource efficiency through
recommending appropriate crops that suit particular features of the soil and
climate. It is a strategic approach that conserves resources, thus avoiding wastage.
2
•Increased Productivity:
Promoting overall agriculture produce will be a major goal for the system. The
system directs farmers towards crops that flourish best in the weather and soils
around them hence high yields. It is expected that this alignment will lead to an
increase in harvest multiples.

•Risk Mitigation:
Many farmers encounter unpredictability on weather issues and soil variation. To
tackle this challenge, our prediction system advises a selection of crops that are
resistant to specific climatic conditions. This risk management approach protects
farmers from negative effects resulting from volatile climatic conditions.

•Cost Savings:
It is indeed crucial as it will help in making farming cost effective. Farmers can
also be advised on which crops are most suitable for cultivating in certain
conditions using inputs like fertilizers, pesticides, and water. Not only does this
lower operating expenses, but it is also important for proper natural recourses
utilization.

•Data-Driven Decision Making:

The system empowers the farmers with actionable insights for better informed
decisions based on data. Farmers need real-time data with which they can move
away from traditional but often non-reliable methods in order to make proper
farming decisions that are more productive and efficient.

•Adaptability to Changing Conditions:

Despite changing weather patterns as well as variable soil conditions, the
prediction system is flexible for any given scenario. Additionally, it constantly
updates crop recommendations to match changing agro-ecological factors, which
provides farmers with appropriate and current advise.

3
•User-Friendly Interface:

A friendly platform is user-oriented and recognizes that every farmer is unique.

The interface is designed specifically tailored to be friendly and easy to navigate
for those who are non-tech savvy.

•Sustainable Agriculture Practices:

However, the general purpose of a prediction system is to develop sustainable

agriculture. It fits into the world effort toward sustainable eco-friendly farming
because it recommends such crop varieties and production methods.

In a nutshell, we bring to you

our project outline-

Data Collection:
One of the vital steps in creating a credible crop advice system is gathering
appropriate data. That implies collecting records for both previous and recent
weather data indicating climatic trends that have occurred over a period of time.
Moreover, specific soil analysis should include

information about the NPK levels, pH and other measurements which will enable
farmers to understand their environment in detail.

Besides providing assistance to farmers, this crop recommendation system also

has a profound impact on other aspects of agriculture. It is worth mentioning that
it enhances sustainability and awareness of the environment. The system
empowers farmers to choose the right crops depending on localized weather
conditions, so that they adopt environmental conservation agriculture.

4
Data Preprocessing:

Finally, following the collection of this data, an elaborate and thorough pre-
processing stage begins. This includes cleaning and preparing the data so that
it will be of good quality and reliable. This entails handling of missing values,
dealing with outliers and finally, normally distributing the numerical ones. In
addition, new data adjustments are introduced with the intention of improving
the predicting ability of this model. They strengthen the general reliability of
the recommendation system.

Model Training:
Training your model is vital as you will want to make your recommendation
system’s heart be strong enough. Here, a hybrid dataset composed of observed
climate data, as well as information on soil, is used to train the selected model.
However, to evaluate the model's performance accurately, this dataset is
judiciously divided into two components: separate dataset; one for training the
model and the other for testing its effectiveness. This split guarantees an overall
look into the ability of the model to generalize and to accurately predict.

Prediction:
Having built an effective model, I can now use it to come up with suggestions of
the best crops for planting. New combinations of weather and soil information are
put into the trained model which gives advice on appropriate crops one can grow.
The data are provided here in a form of science-based practical and easily
implementable advice for the farmers. Prediction is the application stage of what
the model has learned and gives useful information for decision making in
agriculture.

Expanding the Scope: Significance, Challenges, and Potential Impact

5
Agricultural Challenges:

The agricultural industry is facing various issues such as the consequences of

climate change and the demand to feed a more and more people in a sustainable
way. Changes in climate such as unpredictable weather, hot or cold temperatures
and the arrival of new pests and diseases, are the main factors that threaten crop
production and food security. Besides, the world's population is expected to go
beyond 9 billion by 2050, which means that there is a haste to increase agricultural
productivity while at the same time, the environmental degradation and the
resource depletion are minimized.

Interdisciplinary Collaboration:

Bridging the knowledge gaps is the job that connects people and areas that are far
away from each other, hence, explaining why it is needed to create a better future.

It is vital to deal with the multi-faceted problems that are facing agriculture by
using a multidisciplinary approach that brings together the agronomic knowledge
and the latest technologies. The creation of a crop recommendation system
demands the joint work of specialists from various disciplines, such as agronomy,
soil science, meteorology, computer science, and data analytics. The use of
domain-specific knowledge and advanced computational methods can be
combined in order to get the power of data to assist the evidence-based decision-
making in agriculture. The main cause of the enormous economic and social
potential is the long-lasting nature of the energy crisis.

The adoption of a crop recommendation system can be a significant boost to the

economy and the social life of the farming communities in the whole world.
Through the provision of farmers with the personalized advice related to real-time
weather and soil data, we can boost the productivity of the farm, use the resources

6
more efficiently and thus become competitive in the market. Besides, the farmers
who are provided with the cutting-edge technologies are able to develop the rural
areas, improve their lives and thus, they are able to achieve a common growth
which is the development of agricultural value chains.

Participating in the Sustainable Development Goals is the process of offering a

valuable input to all the goals that one has in mind related to the sustainable
development.

The crop recommendation system is in line with the United Nations Sustainable
Development Goals (SDGs), especially those pertaining to zero hunger, climate
action, and sustainable agriculture, and therefore, it is a concrete way to reach the
global development objectives. Through the encouragement of the
environmentally sustainable farming practices, the increase of food security and
the capacity to withstand the climate change impacts, we can make a contribution
to the more equitable and sustainable future for all.

Scaling Impact: Through the process of global adoption, the new fad has spread
worldwide. The scalability and replicability of the crop recommendation system
offer huge prospects for application and influence of the system that are far from
the limits of our original study area. Through knowledge sharing, building
partnerships, and using the digital platforms, we can speed up the spread of the
data-driven agricultural technologies in all the agro-climatic regions of the world.
By working together and being committed to the cause, we can provide the farmers
with the necessary help, strengthen the food systems, and create a more resilient
future for agriculture and society in general.

Creating Technology as the key to the farm becomes a vital part since it helps to
survive in an ever changing world.

7
Farming is a hard job, but a good idea can make it much easier. A possible idea is
to employ the use of computers to assist farmers in selecting the crops they want
to grow. Hence, the students can get the school hours which serve the purpose to
decrease the travel, cutting the cost and at the same time, help the environment.

Understanding Farming Challenges

Agriculture is a very hard field to work in because of the numerous difficulties it

contains. Occasionally the weather is capricious, or there are insects that consume
the plants. Moreover, food resources are increasing but not so much land for
growing more of them.

The saying that "two heads are better than one" is proved by the fact that everyone
in the group is able to come up with solutions to the same problem. The farmers,
scientists, and other people are cooperating with each other to come up with
solutions. They are using computers and special programs to assist farmers in
selecting the appropriate crops that will grow well. Thus, farmers can increase
their crop yield with less problems.

Technology should be the thing that everyone can use without any problem even
if they have no idea how to use it. It's crucial that the farmers can use this
technology with no problem. Computers and sophisticated applications are
frequently difficult to master. Though the new tools are more complicated than the
old ones, they are designed to be easy and friendly for all people to be able to use
effectively for a better yield of crops , since this software is regional specific its
predictive accuracies can be sensitive to data .

8
Learning from Each Other:

Everyone is constantly acquiring new skills and attempting different approaches

to life. It is through the exchange of information and the collaboration of the
farmers that they can make better decisions. Thus, they become more productive,
waste less money, and lessen the damage to the environment. The anticipation of
a good life in the upcoming years is what drives me to live a life. With the
advancements that are in the pipeline, farming could be even better in the future.
Technology and cooperation of one another, farmers can increase the production
of food, consequently, protect the environment, and thus, life will be better for
everyone. Let us collaborate for a superior future!

Conclusion:

In summary, the increasing of our project scope emphasizes the importance,

difficulties, and the possible effect of the application of machine learning
algorithms to provide crop recommendations in agriculture. Precision agriculture
principles, cooperation of different experts, and the global development agenda
are the ways to our future of the agriculture that is sustainable, resilient, and
equitable. The farmers, scientists, and other people are cooperating with each other
to come up with solutions. They are using computers and special programs to assist
farmers in selecting the appropriate crops that will grow well. Thus, farmers can
increase their crop yield with less problems.

The fusion of computers and specialized software with agricultural practices

represents a transformative shift in modern farming. Through sophisticated
algorithms and data analysis, these tools empower farmers to make strategic
decisions that enhance crop yield and quality while minimizing risks and resource
wastage.

9
 Fig 1 Project Outline

1.2 PROBLEM STATEMENT

It is the agricultural sector that acts as a link between food security and economic
stability. Farmers face difficulties when it comes to properly deciding on what
crops to grow thus leading to poor output, wasted resources, and lost money. As a
response to an urgent solution, this project aims at creating an efficient and
accurate crop recommendation system using machine learning techniques.

This anticipated system will help farmers in making crop recommendations taking
into consideration different imperative aspects. A sophisticated recommendation
system that applies modern machine learning technologies in order to provide
farmers with information regarding the optimal choice of crops in terms of their
potential outcome in the agriculture sector is what this refers to. These include
land characteristics such as the quality of soil, average rains, general weather
pattern among others region specifics.

10
This project aims to fill a gap in modern agriculture by integrating ancient
techniques with new technologies. The recommend system will integrate machine
learning algorithms in order to understand complex datasets about soil types,
precipitation rates, among other subtle weather factors in specific regions. This is
where the effectiveness of the system comes in. It integrates this data and gives
out customized recommendations for a specific farming zone.

The proposed crop recommendation system aims at providing an advanced tool

that surpasses merely wisdom to farmers finally. The framework incorporates an
extensive list of factors comprising of the characteristics of soil, rainfall, and
weather prediction in rethinking agriculture decisions. The novel approach aims
at increased yields, resources utilization, and resilient financial base for the
farmers towards enhancing the sustainability of agriculture.

Besides providing assistance to farmers, this crop recommendation system also

has a profound impact on other aspects of agriculture. It is worth mentioning that
it enhances sustainability and awareness of the environment. The system
empowers farmers to choose the right crops depending on localized weather
conditions, so that they adopt environmental conservation agriculture.

In addition, use of machine learning algorithms improves the accuracy of crop

advisories and provides a basis for further agriculture breakthroughs. It shows a
higher tendency of embracing contemporary farming methods that the focus is on
integrating the latest technology. With time, this system will be beneficial in
advancing precision agriculture.

This conforms with other global initiatives that seek smart and sustainability
approaches in agriculture. However, the project is part of a larger discussion
concerning use of technologies for managing growing populations, volatile
climatic conditions, and food safety.
11
scale agricultural landscapes. This could create an ecosystem by which farmers
share information across boundaries that can fill up this system much better for
future.

To conclude, the projected crop suggestion system is not only a response to a

particular farmer’s problem but also the initial point to a brighter, more
sophisticated and linked future for farming. Therefore, this project is an important
step in securing prosperity for every farmer as well as the agriculture sector itself.

1.3 OBJECTIVES

Main Goal:

The major goal of this project is to change the agricultural production for better
productivity. Machine learning technology is used in the recommendation system
to provide advice for crops that are appropriate for each agricultural zone given the
prevailing weather conditions. The crop yields will be improved tremendously,
thereby making the agricultural production more sustainable and efficient.

1. Resource Efficiency:

The main objective of this project is to increase the efficiency of inputs in the
agricultural industry. Specifically, the recommendation system should be used to
conserve these scare resources like water, fertilizers and pesticides. This reduces
the resource wastage as the system actively leads farmers to appropriate crops for
the local conditions. Sustainable farming considers environmental protection and
resource efficiency, thus striving for the environment conservation.

2. Risk Mitigation:

12
As a preventative measure for farmers, it is equally an aid in ensuring informed
and safe decisions are made. Farmers can think of several things like climate
conditions, diseases which can help in their decision making based on a reduced
risk. This method builds resilience in the agronomy sector, a sort of shield against
factors impeding crop production.

3. Sustainability:

The project aims at fostering an environmental-friendly strategy. The

recommendation system directly encourages sustainable agriculture by offering
advisories on crops and resource management. The project is closely connected
with global endeavors for a green agricultural landscape. The project reduces the
utilization of excessive resources hence promoting sustainable farming.

4. Economic Benefits:

However, economically, the project transcends the boundary of individual farms.

The recommendation system has a cascading effect on farmers’ profitability
through directly influencing crop yields, optimizing resource allocation, and
mitigating risks. In this way, it contributes to rural economic growth, which
develops a robust and sustainable agricultural society. This economic effect of the
project echoes throughout the entire supply chain, touches markets, and
contributes to the general economic prosperity of the region.

Summarizing the objectives of the project, they encompass the transformation of

agricultural practice, improving resource efficiency, risk management,
sustainability, and economic development. The project envisages a future where
agriculture is more productive, sustainable, resilient and economically vibrant
through the implementation of an advanced crop recommendation system.

13
1.4. SIGNIFICANCE AND MOTIVATION OF THE PROJECT WORK

The story of prosperity and sustainability is a transforming thread in the complex

art of agriculture. This undertaking is significant at its core transforming the face
of agriculture. To start with it stems from the appreciation that our farmers’ voices
are silent yet there is a lot to do for them. Today, as food demand soars, our
aspiration is to raise agricultural production and to a greater extent.

Our complex machine learning algorithms, help us recommend to our users. This
system is not only a technology invention but a practical method to the daily
problems faced by farmers. Any doubt as to what is at risk is unfounded – to assist
farmers to make informed and precise decisions based on scientific evidence. A
crop improvement is not simply about an increase in yield. It is about building a
resiliency and sustainable crop that the world depends on for food security and
economic stability.

It is also about technological progress, not just the motivation. Efficiency in

agriculture and resource utilization: correct application of water, fertilizers, and
pesticides. This project will provide suggestions of the best crops for the local
condition and minimise resource use as the global voices for sustainable
agriculture become more intense. It’s about the farmer and the inherent risks in
agriculture; the weather, diseases, and fluctuations in market demands. This
shields farmers and has the means to traverse through uncertainties, as they build
on resistance.

Sustainability is not a destination; it is a journey. Our sustainable practice

concentrates on friendly farming, a world agenda for sustainable agriculture. This
is an economic vision that goes beyond a single farm, and extends to the whole
community.

The project assists in rural economic development by enhancing crop yields

through optimization of inputs and minimization of risks to production. It is an all-
14
encompassing approach, imagining a world where each harvest is a celebration of
plenty and every farmer thrives with the necessary resources.

The project is aimed at helping farmers and making sure that there is enough food
for everybody. We apply the smart computer programs to the farmers to inform
them how to grow more food and make the right decisions. Thus, this, in turn, is
the main reason for their growth. Farming is a tough activity due to the many
problems like the bad weather and the bugs that eat the crops. Nonetheless, with
our help, farmers will be more successful and therefore, the food supply will be
enhanced for the entire population.

The purpose of our work is to turn farming into a more beneficial activity for
everybody. We strive to use technology in a manner that the environment will be
protected and farmers will be flourishing. The cooperation and the fitting use of
your thoughts are the keys to the future of farming. Today, everyone knows that
agriculture is a major reason of the environmental stress, and farming should be
managed in a sustainable and prosperous way through cooperation and the smart
thinking. This project is just the beginning of a better future of farming, where
everyone will be able to have many foods and also, will be happy.

1.5 REPORT ORGANIZATION

The project report has been divided in six major parts which contain important
information related to the topic.

CHAPTER 1: INTRODUCTION

This first chapter constitutes the project’s launchpad where a specific problem is
put forward, the intended objectives identified, and reasons for undertaking the
venture explained.

15
CHAPTER 2: LITERATURE SURVEY

This is a part of chapter where we dig up knowledge that has already been there,
using good materials like technical papers, books etc. We want to understand how
our market looks at present and what we can address in our project.

CHAPTER 3: SYSTEM DEVELOPMENT

Here, in this chapter I shall cover the essentials on the project, from requirement
analysis to the system design and implementation. Talking about challenges faced
after development and strategic repair.

CHAPTER 4: TESTING

The following section explains our testing strategy and procedures that were very
thorough. We provide test scenarios together with results that reveal the
consistency of the system very clear.

CHAPTER 5: RESULTS AND EVALUATION

Herein lies the concluding chapter whose focus is on results and examines the
outcomes as well as compares them with existing remedies. It provides in-depth
analysis of our work.

CHAPTER 6: CONCLUSIONS AND FUTURE SCOPE

Concluding our research development and working on improving the project by

adding more functionality to widen its scope for future scope.

16
 CHAPTER-2 LITERATURE SURVEY

In 2021, Mohammed Adnan together with other developed a very crucial research
about “a machine learning based crop recommendation system based on soil and
weather data“[1]. Therefore, the authors proposed several suggestions to farmers
on what crops should be planted in response to challenges presented by climate
change and pressure of the growing population.

The methodology was comprehensive and included data collection, clean-up

analysis, choosing appropriate modelling approach, as well as ongoing
improvement process. One of the milestones towards the project was the creation
of a crop recommender system that used several algorithms like Decision tree,
KNN, SVM and naïve bayes which resulted in exceeding 98 percent of accuracy

This article also recognized their successes in addition to stating the challenges
that must be overcome such that future prediction of agro-based technologies
can be enhanced. It highlighted two major points which can have negative impact
on them namely, data quality and selection of algorithms. In addition, this article
will be important not only for modern view on precision agriculture, but will also
emphasize that we need to make our further steps towards satisfying requirements
of constant update.

In addition to it they found the limitations in the research due to crop diversity and
crop [Link] rotation is an organized strategy for planning farming in which
various crops are planted in the same plot of land in alternate seasons or years.
This refers to switching up crop types in certain areas at an interval. Crop rotations
are applied for improving soil health, raising fertility levels, and eradicating pest
and disease problems.

By definition, agricultural data are variational in nature, for reasons associated

17
with various environmental dynamics. It is essential to ensure that the data
collected are diverse so as they can represent various types of soils, weather
conditions, and modes of practice under Agriculture. Nevertheless, getting this
dataset having adequate variation is problematic since some regions or particular
plants can be overlooked.

Crop recommendations system involves one major setback pertaining to

availability and accuracy of data used for training and modelling. At present, there
are technologies that make it possible to obtain large volumes of agro-statistics;
however, its accuracy, verisimilitude, and completeness is still challenged issue.

According to CVSM practice, temperature, type of soil, humidity and rainfall are
key in determining the amount of crops. The model is based on the use of ANN
and it determines anticipated yields, which inform farmers about the most
appropriate crops they should grow. This study acknowledges the complexity of
the agricultural ecosystem by factoring economic considerations (market prices)
in selecting their crops.

The Crop Vegetation Soil Model (CVSM) is the one that considers a lot of factors
when the farmers are deciding what crop to plant. These factors are the
temperature, the type of soil, the humidity and the rainfall which play the role of
the growth of the plants. The factors that have been investigated by the model
allow it to predict the amount of food that will be harvested in a certain area. It
uses an Artificial Neural Network (ANN) tool to perform this task.

Hence, it lets the farmers to figure out what crops will flourish in their area and
the ones which will not. On the other hand, the model's job is not finished yet. As
a result, it also takes into account the economic side, for example, how much
money farmers can generate by selling their produce. Through the evaluation of
all the above mentioned factors, the model helps the farmers to make the right
choice of the crops to be planted so as to have a good harvest.

18
 Fig 2 Model Architecture 1

The crop variety selection method by G. Vishwa, J. Venkatesh and C. Geetha in

2020[2] was groundbreaking because it enabled application of machine learning
methods for enhancing food production. The complex decision-making procedure
on crop selection that CVSM tackles comes with increasing complications from
unstable natural environment as seen in the agricultural industry. The research
highlights the importance of appropriate choice of crops as a way of minimizing
risk and enhancing profitability of farms.

19
The core of CVSM lies in its three-part algorithm: factors such as the type of farm
crops, pricing in the market, and crop variety. The model further measures success
by comparing projected yield production for a certain amount of money and
assessing whether it is profitable depending on existing market prices.

The study shows how artificial intelligence could be applied to replace

conventional farming methods. CVSM is a comprehensive method of making crop
choices that provides a farmer with a system for dealing with unpredictable factors
of environment and to increase crops’ yield. This highlights the need for utilizing
novel technologies to advance agriculture towards sustainability and higher
economic returns.

Their model’s architecture is as follows-

Fig 3 Model Architecture 2

The work, “Crop Recommendation System For Precision Agriculture”[3]

developed a crop recommendation system that uses big data in contemporary
farming practices in India. Here, it is essential to indicate the following big data
types which are crucial and appropriate for
20
predicting and improving of choosing the crops against However, the authors
emphasize on precision agriculture and provide some appropriate yield prediction
models for that purpose.

However, literature review on related work is highlighted on the importance of

crop yield forecasting, environment, and machine learning in agricultural
practices. Four major processes that comprise a particular methodology are data
pre-processing, weather data processing on MapRed, formulating a dataset,
establishing the space distribution between close stations weather data for weather
forecast.

Considerations such as the nature of soil, normal seasonal climatic conditions in

different regions among others would be considered when planning different crops
for planting. Employing a map/reduce style of the two step approach, authors
process the large data sets. Application of nearest neighbour approach and weather
similarity method improves the accuracy for forecasting future crops.

The last issue highlighted focuses on big data. Perhaps this would instill some
level of intelligence in the agriculture sector towards a more intelligent and data
oriented sector. It gives advices to farmers about the way they should determine
their crops depending on past information for the purpose of food assurance in the
country. Authors finally end with the need to identify more appropriate traits for
yield prediction and improved datasets. In general, this study highlights the
significance of precision farming and big data analytics.

The problem of big data in agriculture is significant as it could make the farming
sector more knowledgeable and data-centered. Through the application of big
data, farmers can achieve better decision-making about the choice of crops to plant
after studying the past data. Thus, it safeguards that there is plenty of food to be
consumed by the country. Nevertheless, the authors also stress that much more is
yet to be achieved.

21
 Fig 4 System Methodology

The paper titled "Towards an Efficient Recommender Systems in Smart

Agriculture: Deep reinforcement learning” approach [4], M. Bouni et al., Procedia
Computer Science, vol. A study on a DRL-based recommender system in
precision agriculture takes an insight into the literature (Zhang 2022, p. 203, pp.
825–830). In this regard, the study contrasts the new DRL approach against
traditional machine learning models including Random Tree, Naive Bayes, and K-
Nearest Neighbor for making precise crop yield prediction required by farmers.
The paper mentions about possible use of DRL, emphasizing that it is likely to
perform even better than random forest if applied on large quantities of data.
However, no details are provided regarding implementation challenges or future
improvements. Summarily, exploiting modern technology within the agriculture
decision making process is vital towards realizing an improved DRL based crop
recommendations model.

The paper, "Towards an Efficient Recommender Systems in Smart

Agriculture”[5]: A precision agriculture recommendation system based on deep
reinforcement learning approach has been discussed in one of their publication
papers titled “A Deep Reinforcement Learning Approach” by M. Bouni et al.
22
Yield prediction like random tree, naïve bayes, and k-nearest neighbours. The
paper does not, however, give any information regarding the implementation
problems or future improvements with regard to DRL even though it accepts the
possibility that DRL could be better especially for larger dataset. In summary, this
study highlights that incorporating modern technology in decision-making about
farming is crucial for enhancing DRLO-based crop advisory techniques.

2.1 OVERVIEW OF RELEVANT LITERATURE

Nine notable literature reviews on “crop recommendation” cover machine learning

in agronomic recommendations. Many of these surveys do an extremely detailed
analysis of the strengths and weaknesses of crop recommendation systems. The
effect of climate change, lack of available long dated data, variable soils, and
rotation in the growth of crops has been studied by researchers. These surveys
show that, while predictive models may focus on specifics like pests, diseases,
resources, or technologies as recurring themes, they can also be affected by
broader concerns such as human factors. This aspect has also focused on
agriculture’s infrastructure problems. Taken together, these literature surveys
provide a comprehensive overview of the issues involved into developing crop
recommendation system and give new perspective into this project

•Mohammed Adnan & Patel, A. (2022)[16]:

Conclusion: Underlined the significant importance of joint work between the

research companies and those responsible for developing crop recommendations
in large scale.

23
•Gandhi, R., et al. (2022)[17]:

Conclusion: The study found that adoption of crop recommendation mobi-

applications facilitated accessibility among the farmers.
Key Insights: Discussed on issues regarding digital literacy and availability of
connectivity in remote areas with proposed solutions to addressing these hurdles.

•Nath, P., & Meena, M. (2021)[18]:

Conclusion: Henceforth, it concluded that the consideration of economic viability

makes recommended crops practical in use and improves effectiveness of
recommendation systems.
Key Insights: Dealing with market uncertainties and incorporation of a dynamic
model in the decision making process.

•Bansal, S., et al. (2020)[19]:

Conclusion: They concluded that sophisticated data visualization enhances the

accessibility and understanding of crop recommendations insights.
Key Insights: Discussed communication of uncertainties among predictions and
ways to increase farmers’ confidence on the system.

•Jha, A., et al. (2019)[20]:

Conclusion: Finally, concluded that it is important to ensure the scalability of crop

recommendation models so that they can be adopted in different types of
agricultural landscapes. Key Insights: Detailed on the issues of model transfer
between regions, and how to achieve high precision in new circumstances.

•Singh, R., et al. (2018)[21]:

24
Conclusion: It was concluded that the use of soil health monitoring
technologies in formulation of site specific and personalized crop
advisory services is crucial.
Key Insights: Challenges on soil variability and precision approach in
recommendation scheme.

•Chen, J., et al. (2018)[22]:

Conclusion: It ended up concluding that a participatory strategy which involves
the farmer in the model training yields more precise, and appropriate
recommendations.
Key Insights: Identified issues of farmers’ suspicion and suggested involving
designers in the process of making a product simpler.

•Wu, H., & Li, X. (2017)[23]:

Conclusion: To this end, I have concluded that considering the environmental
implications of crop selections is imperative for an environmentally conscious
agriculture.
Key Insights: Challenges associated with ecological balance and a
comprehensive approach toward including environmental characteristics in the
recommendation algorithm.
Gupta, S., et al. (2021):

Conclusion: Ultimately, concluded that it is imperative to conduct continuous

model monitoring and adaptation for crop recommendations that are stable and
reliable over extended periods.
Key Insights: Provided the addressed troubles associated with the changing
agriculture practises along with the feedback loop arrangement for systematic
enhancement of the approach in question on a continuous basis.

2.2 KEY GAPS IN LITERATURE

25
Limited Integration of IoT and Advanced Technologies:
There exists a notable lack of focus on the use of IoT and other sophisticated
technologies in enhancing crop recommender systems. Some researches hint at
data acquisition through IoT but there is very little comprehensive study about
utilizing it for real time-tracking, precise farming, and sensor insights. By
incorporating IoT within the environment, the data collected can be accurate and
timely so as to build on new age basis for recommendation models.

Detailed Explanation:
With the merging of IoT, sensors, and modern technologies, the agriculture sector
is undergoing a very fast transformation. Nevertheless, the literature on crop
recommendation systems does not fully address how IoT can impact issues related
to crop advisory systems. To fill in the above-mentioned gap, comprehensive
studies investigating the use of IoT for continuous monitoring, feedback loops,
and precision agriculture approaches must be conducted.

Insufficient Consideration of Socio-economic Variables:

Most of the literature survey and researches emphasized mostly on environmental
and agronomics aspects while ignoring socio-economic parameters that are vital
in the decision making process for farmer. However, poor focus on socio-
economic considerations including market trends, economic pliability and social
dynamics gives rise to this gap towards comprehension of total environmental
circumstance.

Detailed Explanation:
Apart from soil fertility, weather patterns and other natural endowments,
agriculture depends on various economic and social factors of production. Most
literatures fail to provide detailed investigations as to the influence socio-
economic aspects have on crop selection. Addressing

26
this disparity will come through the development of economically conscious
agronomic decision-support tools with more realistic and practical guidelines for
farmers’ practices.

Scarcity of Studies on Crop Diversity and Regional Specificity:

Most of the old literature concentrates on main crops leaving out a wide range of
others grown in different places. Sustainable farming and varied crop types are
critical issues that need addressing in the light of recommendation models that
address farmer’s real needs in various agro ecological zones.

Detailed Explanation:
Agriculture manifests in varied forms, as farmers grow different crops depending
on the regional environmental factors and consumer preferences. Regrettably
though, such studies on recommendation model adaptation are rare in the
literature. This gap can be addressed by undertaking research which takes into
account regional specificity considering the specific difficulties and advantages of
distinct crops in various regions.

Inadequate Exploration of Climate Change Impacts:

Agriculture being affected by changing climatic conditions, there is not much
written material that has been used to include climate changes factors into
recommended crop model. A delicate perspective on the impact of changing
climate parameters on crop suitability has been largely absent in the extant
literature.

Detailed Explanation:
Uncertainties are created by climate change about weather patterns which can
greatly affect the ability of plants and crops to grow properly. The detailed
empirical studies of potential crop recommendation systems are usually missing
in the literature. Filling such a gap will necessitate consideration of the link
between integrated climate change models, predictive analytics, and adaptive

27
strategies and increased crop recommendation system resilience.

Addressing such important loopholes in literature is a significant step towards

improving crop recommendation systems. However, a more holistic appreciation
encompassing cutting-edge technologies, socio-economic factors, crop diversity
and climate change needs to be considered when building sustainable agriculture-
driven recommendation systems.

5. Limited Emphasis on User-Centric Design and Adoption:

However, a major lack within the existing literature relates to little consideration
being placed on the implementation of farmer’s crop recommendation system
following user-centered design approach. In spite of numerous studies addressing
the technical aspects associated with modelling, there remains little literature
exploring issues related to usability, user interaction, and sociotechnical factors
which significantly impact the acceptance and implementation.

Detailed Explanation:
User acceptance or adoption is critical in any technology involving crop
recommendations system. Most of the present research focus more on the
algorithm complexities and data analytics as opposed to the user-centric designs.
This calls for research on what farmers’ opinions concerning adoption of digital
technologies, preferences and the problems they face. It is important to understand
the socio-technical context as well as include user feedback during the design
process when developing effective and user-oriented crop recommendation
systems.

28
 CHAPTER-3 SYSTEM DEVELOPMENT

3.1. REQUIREMENTS AND ANALYSIS

• SOFTWARE RESOURCES

● Python(Version: 2.7)
● Seaborn
● Matplotlib
● MaxMin Scalar
● NumPy
● Pandas
● Scikit-learn

● HARDWARE RESOURCES

● 8 GB RAM
● GPU

3.2. PROJECT DESIGN AND ARCHITECTURE

Architecture of crop prediction system based on meteorological information is

extremely elaborate in order to facilitate an integrated, effective, and intelligent
solution. It involves various interconnected modules and components:
• Data Collection Module:
Objective: Crucial data inputs for crop prediction such as real time
meteorological data, historical crop data, and soil characterization.
Implementation: It gets live updates through APIs about the weather, pulls
historical data from agricultural database and accesses soil databases for
relevant information about the soil.

29
 • Preprocessing Module:
Objective: Addresses missing values, ensures consistency across
measurements (e.g., in relation to questionnaires) etc.
Implementation: Uses sophisticated data cleaning algorithms, approaches to
handling missing data, and normalizations for numerical characteristics.

• Extracting Useful content:

Objective: Enhances the predictive capability of the machine learning model
by extracting features that are useful from the dataset.

Implementation: Uses complex algo that are good at identifying various

features.

e.g. changes in temperatures, rainfall amounts, or nutrient availability in soils.

Creates features by leveraging domain knowledge.

•Machine Learning Model Module:

Objective: Precess highly processed dataset of development, assessment, and
validation of different machine learning models to achieve highly precise
prediction of crops.
Implementation: Carries out comprehensive tests with various models, such as
neural networks, using PyTorch (version 2.0.0) or similar toolkit. It is a tedious
procedure entailing testing on historic information and subsequent assessment.
The aim is to determine the best and credible model, which can be able to assist
them in successful predictions.

•User Interface Development:

Objective: Tries to create an interface that is easy to use and enables the farmers
to have smooth interactions with the system. This enables users to modify input
variables promptly and show forecasts. Implementation: Developing a functional
user-friendly website that understands the language of the users using the MERN

30
(MongoDb Expression, React, Nodejs) language stack. These web based
interfaces developed using frameworks such as react for dynamic interactions
enable reachability on behalf of farmers. This allows for easy adjustment of the
parameters such that the crop recommendations from the website are quite
understandable.

•Integration Points:

Data Flow: It provides seamless passage of data between collection module,

through preprocessing, feature engineering and ultimately towards machine
learning model.
Module Integration: Provides smooth synching between the ML model and the
feauture engineering sub-module that receives processed variables for more
accurate predictions).

•Scalability and Flexibility:

Scalability: Scalable in handling different datasets sizes prepared for possible

future expand on data volume and/or other functionalities.

Flexibility: Facilitates other machine learning algorithms or modifications to

machine learning algorithm in later iterations.

•Security Considerations:
Data Security: Makes use of strict measures of access control and encryption
especially involving farmer specific information, if the data is sensitive.

•Performance Optimization:

Parallel Processing: Speeds up data processing by optimizing

performance via parallel processing techniques.

31
Algorithm Efficiency: It uses very efficient algorithms and therefore minimizes
processing time while maximizing prediction accuracy.

•Maintenance and Updates:

Modular Design: Modular architecture makes it possible to update each module or
part individually and without affecting the operation as a whole.

Version Control: Strong revision control system that traces all modifications.

Essential modules are interconnected in such a way to produce an overall

intelligent crop prediction system’s architecture. The system starts with strong
data acquisition and pre- processing where it is able to pick only significant
features for high performance machine leak model.. ML module vigorously
evaluates different models and ensures high accuracy in crop prediction. The
initial approach of user interface development through a MERN based website
prioritize friendly interactions with farmers. Seamless data flow is accomplished
through integration points, while scalability and flexibility allow for adaptation to
new datasets. The security of confidential information is ensured whilst
performance enhancement ensure the proper processing. Modular configuration
allows for simple repairs and upgrades, indicating an agile and adaptive system
for predicting agriculture.

The system of different parts integrated into each other is what will make the data
flow seamless and thus, the processing and the use of the data will be efficient in
the whole work. The data flow mechanism is the main tool in the data collection
process that coordinates the transfer of information from the data collection
module, through the preprocessing stages, feature engineering, and in the end into
the machine learning model. This perfectly scheduled process of aggregation of
data enables the system to draw out the meaningful information and give the right
predictions.

32
The first and foremost important part of integration is the alignment between
different modules, especially the machine learning model and the feature
engineering sub-module. This synchronization makes it possible for the cross refer
to the processed variables, which then can be used for the model to get better
prediction. Through the guarantee of the fluidity of the communication and
interaction among these components, the system improves its predictive capacity,
thus, enhancing in the performance and the accuracy.
means the system can accommodate both the growth of the data volume and the
new functions in the future without any problems. Hence, the full scalability of the
system makes sure that it remains efficient.

The Data flow diagram is as follows-

Fig5 Flowchart of working of our model

33
3.3 DATA PREPRATION

Source of Data:

The data set exclusively obtained from Kaggle covers every detail regarding soil
composition, comprehensive weather report, among others, while also considering
the effect precipitation may have on plant growth.

• Soil Content Dataset:

This dataset comprises three main components:

Nitrogen (N): Crucial for plant growth and development as it is a measure of the
amount of nitrogen contained in the soil.
Phosphorus (P): Reflects on the state of phosphorous present in the soils, which is
needed for plant’s energy transfers.
Potassium (K): Refers to the level of potassium in the soil that is essential for the
whole-body strength of the plant.

• Weather Report Dataset:

Meteorological variables that are essential in agricultural decision making fall
under the weather report dataset.
Rainfall Data: Provides detailed information on pattern of rain in terms of crop
productiveness. Impact of Rainfall on Crops:
The data set looks at the relationship between rainfall and crop yields, an
important aspect of agriculture which has been affected by the changing nature of
weather patterns.

• Data Structure:
The dataset structure includes distinct CSV files:
Main Kaggle Dataset: Each item contains specific soil component descriptions,
weather information, as well as data about how much rain impacted them.

34
Additional Dataset Features: This includes information regarding how rain affects
crop growth.

• Dataset Reference:

This project’s dataset borrows its idea from the Kaggle platform, where it offers
unique information on NPK content of soil, multiple weather indices and specific
focus on effect of precipitation in relation to crops growth.

3.4 IMPLEMENTATION

● DATA COLLECTION AND PREPROCESSING:

• Data Collection:
Source:
All the data employed within this project was obtained exclusively from Kaggle;
this encompasses very essential data needed within a Crops’ recommendation
system. Nitrogen (N), Phosphorus (P), Potassium (K), temperature, humidity, pH,
and rainfall are some of the data points in the sample set.

Dataset Labels:
• Nitrogen (N):

It is a measure of nitrogen in the soil, one of the components affecting vegetation growth.
• Phosphorus (P):
Means of energy transfer to plants within the soil indicating its phosphorous level.
• Potassium (K):
Means potassium content of the soil which is essential for plant’s
overall health and toughness.x
• Temperature:
35
 Environmental factor – ambient recorded temperature which is critical in plant growth.
• Humidity:
The amount of humidity captured is very significant as it influences the
plant’s transpiration and growth.
• pH:

Indicates soil acidity / alkalinity, which is an important consideration in

terms of nutrition availability to plants.

Rainfall:
The amount of rainfall influences crop productivity; therefore, this provides
information on rainfall level.

Data Preprocessing:

CSV Format:
Obtained in CSV for the Kaggle made it manageable and straightforward to
incorporate it in machine learning worksheets.
• Preprocessing Steps:
Handling Missing Values:
Filled up and addressed any missing values in the dataset for data integrity.
• Normalization:
Used normalization methods to normalize numerical features range to ensure uniform
dataset
for training machine learning model.
• Feature Engineering:
Adding new additional features or transforming previous to increase the model’s
capture meaningful pattern.
• Label Encoding:
Label encoding for machine-learnable converted categorical data into numeric format.

36
Fig 6 Dataset

Fig 7 Soil Content

37
Fig 8 Dataset Visualization

38
Fig 9 Data Preprocessing

Fig 10 Soil Data

39
 Fig 11 Preprocessing

MODEL

During the process of systematically applying some selected machine learning

algorithms towards empowering the Crop Recommendation system to predict crop
yield and recommend corresponding crops accurately and reliably by depending
on agricultural parameters within the data set involved.

K-Nearest Neighbors (KNN):

Using the K nearest neighbor algorithm as a non-parametric classifier, we were
able to quantify similarity between input parameters and their historical analogues
in order to sort crop types. KNN considers that objects similar in the feature-space
have characteristic properties in common and are thus likely to be of like-class.
KNN classification involves measuring the distance between data points in order
to recommend crops appropriately.

40
Fig 12 Model Selection

41
 Fig 13 Model Selection

Fig 14 Histogram for Nitrogen Values

42
Support Vector Classifier (SVC):
An SVC helped construct hyperplanes in the feature space that distinguished
between various crops. SVC involves locating the best separation plane on which
two different classes can be reliably classified. This algorithm proves useful for
managing complicated datasets characterized by non-linearity among the
individual attributes or features.

Random Forest:
We employed the Random Forest algorithm as an essential part of our ensemble
learning in order to enhance our model’s prediction accuracy and robustness. The
Random Forest generates a lot of decision trees during training and then provides
as output the majority of the individual trees for classification. Such ensemble
approach reduces over fitting which ensures good generalization performance
suitable for Crop Recommendation system.

Through using Pickle, we gained the ability to serialize and deserialize Python
objects, thus, we could save the trained machine learning models in an efficient
way. This prediction system takes into account of the NPK values as the input and
uses the pre-trained model to predict the best crops for the given nutrient
composition.

The essence of employing the Random Forest algorithm lies in its ability to
harness the collective wisdom of numerous decision trees. By generating a
multitude of decision trees during training and aggregating their outputs through
a voting mechanism, Random Forests mitigate the risk of overfitting, thereby
enhancing the model's prediction accuracy and robustness. This ensemble
approach enables the model to capture diverse patterns and variations in the data,
leading to improved generalization performance. In the context of a Crop
Recommendation system, where the goal is to provide tailored recommendations
based on complex agricultural datasets, the Random Forest algorithm's ability to
handle high-dimensional data and nonlinear relationships makes it particularly
43
 well-suited for the task.

Fig 15 SVC

Furthermore, SVM's ability to handle high-dimensional data and its flexibility in

accommodating nonlinear relationships make it a valuable tool for complex agricultural
datasets. Through careful feature selection and parameter tuning, SVM can offer accurate
recommendations tailored to specific agricultural contexts.

In summary, the incorporation of Support Vector Machines in crop recommendation

systems enhances decision-making processes for farmers and agricultural stakeholders,
leading to improved crop productivity and sustainable farming practices.

44
 Fig 16 Random Forest

Decision Tree:
By using some selected features for recursively splitting of data in successive
levels, they classified the crops hierarchically via the Decision Tree algorithm.
Decision trees that capture complex relationships with the data for explaining why
the decision is made have been adopted by the Crop Recommendation
[Link] partitioning of the split dataset using Decision Trees led to
hierarchical classification of the crops. The interpretability of the Decision Trees
makes it a useful component of the Crop Recommendation System as they are able
to capture the non-linear relationships between variables in the data.

All the above algorithms were trained thoroughly on the dataset, hence learning
patterns and relationships between crop recommendations and input parameters.
A subsequent evaluation on a testing set provided a comprehensive analysis on
their performance

FINE-TUNING

Immediate Goal: Enhancing Algorithm Performance

At this critical stage, we want first to refine the machine learning algorithms used
in our system of Crop Recommendation namely the K—Nearest Neighbors
(KNN), Support Vector Classifier
45
(SVC), Random Forest, and Decision Tree. With this goal in mind, these actions
refine our approach.

Insights from Evaluation:Based on these evaluation results, we adapt our approach

to target those improvements areas for more accurate prediction of crop suitability
through various agro- parameters.

Hyperparameter Optimization:We carry out a very detailed adjustment of

hyperparameters in order to improve accuracy of our models. This entails fine-
tuning essential parameters such as learning rates or regularization terms in order
to enhance the accuracy of the models.

Adaptive Model Architecture:We adapt the structure of the models based on the
challenges observed in the evaluation process. This approach is adaptive and thus
the algorithms deal with the agricultural peculiarities.

Targeted Enhancements:Our polishing, if you will, is laser directed at particular

elements that the evaluation has pinpointed. This strategy is meant to improve the
algorithms performance and reliability in crop recommendation.

Realism in Recommendations:It is important to guarantee that the suggestions

generated will be relevant to multiple agricultural input parameters such as soil,
weather, among others for the purposes of practical and credible advice for
[Link] Across Circumstances:This iterative approach allows models
to be adaptable to various agricultural input circumstances. Adaptability is
important in order to make accurate recommendations for soils, weather, and
dynamic agricultural [Link] Iterative Refinement:The process
involves a continuous iterative refinement. Adjustments made in each cycle are
guided by feedback from the previous one to provide a dynamic and refined Crop
Recommendation System.

46
By these steps, we incorporate the concept of fine-tuning into algorithm
optimization procedure for the sake of thorough and cyclical approximation
towards accurate and reliable crop recommendations.

DEPLOYMENT

Our model will be deployed on a web app:Our crop recommendation system will
be refined through rigorous testing and fine-tuning to achieve a state-of-the-art
accuracy and will then be deployed on a user-friendly web application.

User-Friendly Interface:The web application will have a user interface with a

simple layout for the crop recommendation model. The design is for farmer user
interaction; it is not technological complex.

Cross-Platform Accessibility:Convenience will be ensured since the online

application will be optimized to ensure it will work efficiently on multiple
platforms and devices like computers, laptops, tablets, and smartphones.

3.5 KEY CHALLENGES

It presents several challenges to be overcome with innovative measures that must

be considered in the development of a sophisticated crop recommendation system
using machine learning and soil and weather parameters as fundamental data.

•Data Quality and Diversity:

The quality and variety of available datasets are one of the main obstacles. It is
also difficult to obtain detailed information on the soil compositions such as N, P,
K, among others, as well as

47
upto date weather reports. Data reliability is also an inherent problem arising from
the fact that there exist different types of soil and weather conditions from region
to region.

•Feature Engineering:
It is very important to choose suitable features from the complex dataset.
Determining which soil and weather parameters significantly influence crop
growth and yield is a function for which domain expertise is required. Feature
engineering also comes with the added complication of missing or inconsistent
data.

•Model Generalization:
Developing a machine learning model with good generalization across
geographical areas and different climate conditions is a big hurdle. System needs
to change with variations in composition of soil and weather for accurate
recommendation universally.

•Dynamic Nature of Agriculture:

Soil health and weather vary dynamically with agricultural practices and
environmental conditions over a period of time. For the model to remain accurate
it must adapt itself to the changes that are occurring in the agricultural scenarios.

•Limited Dataset Availability:

Accurate machine learning requires access to large, quality data sets. Despite this,
in the perspective of agriculture, getting huge datasets with accurate soil and
weather data for various crops and regions may be difficult.

•Interpretability and Explainability:

It is important to make sure that the prescriptive advice offered by system can be
interpreted and understood by these farmers.

48
 CHAPTER-4 TESTING

4.1 TESTING STRATEGY

Testing Strategy for Crop Recommendation System:

Reliability and accuracy of the machine learning model is crucial in the Crop
Recommendation System. The testing strategy involves different stages to
authenticate the functionality, performance and generalization ability of the system.

•Data Preprocessing Testing:

Objective: Ensure that data preprocessing steps were done correctly.

Activities:Examine soil and weather datasets for missing values and outliers.

Ensure raw data converts into the correct model training format.

Ensure consistency in data encodings and scalings.

•Model Training and Validation Testing:

Objective: Analyze the efficiency of the machine learning model.

Activities:Train the model on a fraction of the dataset and verify on another

fraction. Carry out cross-validations to evaluate generalization.

Measure train and validate accuracy to detect overfitting, or underfitting.

•Algorithm Comparison Testing:

Objective: Assess the various machine learning algorithms.

Activities:Use the K-Nearest Neighbors (KNN), Support Vector Classifier (SVC),

Random Forest, and Decision Tree algorithms.

49
 Fig 17 Max Min Sclaer

Using th Max min algorithm as a non-parametric classifier, we were able to

quantify similarity between input parameters and their historical analogues in
order to sort crop types. KNN considers that objects similar in the feature-space
have characteristic properties in common and are thus likely to be of like-class.
KNN classification involves measuring the distance between data points in order
to recommend crops appropriately.

50
•Hyperparameter Tuning Testing:

Objective: Improve model performance by optimizing hyperparameters.

Activities:Use grid search or random search to find the optimal hyperparameter
values. Study the effect of changes in hyperparameters on model accuracy.

• User Interface Testing:

Objective: Make sure that the user interface is suitable

and functional. Activities:Validate user inputs for soil
and weather [Link] users check the validity
of the predictions shown. Evaluate the overall usability
and interface speed.\

4.2 TEST CASES AND OUTCOME

In the first step of our testing strategy, we concentrate on the authenticity of the
data. First of all, we ensure that there is no missing value for both soil and weather
datasets so as to supply the data of full scale. We normalize the soil nutrient
values(N,P,K) that ought to be 0 – 1. We also conduct quality control for outlier
detection in temperature, humidity, pH and rainfall data and the expected outcome
is outliers removal.

We partition the model dataset into training and validation subsets in the process.
Our objective is to have a perfectly trained model without any errors and a
51
reasonable validationaccuracy. We additionally use cross-validation techniques to
ensure that our model is consistent across different parts of the data set.

Secondly, various machine learning algorithms are compared in terms of their

relative performances. Therefore, we will be measuring the K-Nearest Neighbors
(KNN) algorithm’s accuracy, precision, recall and F1 Score. This assessment also
includes metrics for model performance for Support Vector Classifier (SVC) and
Random Forest algorithms. The Decision Tree also gets tested for measuring its
effect in classification using recursive splitting.

Fine-tuning of hyperparameters is critical towards the improvement of our

model’s accuracy. Grid search is employed and the hyperparameters values are
selected to improve the model’s accuracy. Finally, we look at the impact of
hyperparameter changes on the model, we expect improved performance or no
change at all. Data integrity, model training, algorithm comparison, and
hyperparameter tuning for a tested and accurate crop recommendation system.

Besides, we are also interested in discovering the best algorithms by the way of a
whole review of all the important indicators. The sentence above, we go through
the various aspects of the K-Nearest Neighbors (KNN) algorithm's accuracy,
precision, recall, and F1 Score, which are its accuracy, precision, recall, and F1
Score. This article also reviews the other famous algorithms like the Support
Vector Classifier (SVC) and Random Forest that have been tested intensively in
our model.

Besides, our assessment also contains the Decision Tree algorithm which is the
main focus of our study, where the classification power is assessed by the means
52
of the recursive splitting process. The inspection of the details allows us to know
the strengths and weaknesses of each algorithm, therefore, the system will be able
to recommend the best [Link], we do the thorough research on the effects
of the hyperparameter changes on the model performance.

After the performance metrics are checked again after the tweaking, we will be
able to observe the improvements in accuracy, precision, and recall. On the other
hand, when there are no or only slight improvements, we investigate the factors
that affect the model's behavior, hence, we are able to make decisions on how to
improve the [Link], our actions are described as the total approach that
covers the various aspects of ensuring data quality, the powerful model.

53
 CHAPTER-5 RESULTS AND EVALUATION

5.1. RESULTS

Coming to the last stage , the part of refining the data, training the model, and
carefully testing different ML algorithms in our Crop Recommendation System will
reveal all of the outcomes that are crucial to the system. Precision has been the
watchword of the trip and each step of the way taken has been designed to enhance
the forecasting abilities of the system.

In the journey of our model training, we covered some of the notable algorithms—
K-Nearest Neighbors (KNN), Support Vector Classifier (SVC), Random Forest,
and Decision Tree—and each one had its signature on the canvas of predictive
analytics. Similar to skilled artisans,these algorithms sculpted insights from our
dataset, outlining the fine lines that govern the relationship between agricultural
parameters and crop suggestions.

KNN, in this grand symphony of analytics, proved to be the climax, registering an

accuracy score of 0.98. Its approach, based on the discrimination of trends from
past records, laid grounds for accuracy. Immediately after, the Support Vector
Classifier (SVC) showed its strength with accuracy of 0.9745, using hyperplanes
to move across the multidimensional space of classification of [Link]
Forest and Decision Tree algorithms scored 0.9727 and 0.9726 respectively,
proving their ability to contribute to the predictive ensemble.

54
Decision Tree and Random Forest algorithms are the ones that offer the interpretability and
classification as a group of models respectively. Decision Trees are able to reveal the significance
of the features and the way of the decision-making process, which is helpful in the understanding
of the data structure. Through the combination of several trees, the Random Forest boosts the
correctness and the resistance to overfitting, thus becoming a big help to the crop recommending
systems.

The ensemble methods like Bagging, AdaBoost and Gradient Boosting use the collective wisdom
of different models that prevents the errors and biases of models. These methods are very good at
handling the relationship in data which, in turn, makes the crop recommendations more reliable
and more effective. The skill of improving the prediction of crop yields makes them the most
needed auxiliary parts in crop recommendation systems.

With the conclusion the spotlight remains on KNN, our luminary. As far as agricultural decision-
making goes, it is rather far-reaching than just about numerical accuracies. It is the paradigm shift
where data-driven insights make informed choices. The orchestration of algorithms culminates in
a resounding affirmation: KNN is the leading torch of precision showing the way toward the future
in which technology is combined with agriculture.

K-Nearest Neighbors (KNN) is considered effective for crop prediction due to several
factors:Simplicity and Intuitiveness:

KNN is one of the simplest algorithms relying on proximity. The fact that it is simple allows one
to grasp and interpret it, something which is vital for agricultural applications that will be done by
stakeholders who are not well informed about machine learning such as farmers

Crop recommendation systems are all about the precision and the accuracy in order to assist
farmers in the proper decision making. K-Nearest Neighbors (KNN) algorithm is a top contender
that shows itself to be a good choice, especially when you take into account NPK (Nitrogen,

55
Phosphorus, Potassium) values and humidity. KNN does not care about proximity and hence it is
a perfect algorithm for tasks where the spatial relationships are the key. Through the detection of
the nearby data points having the same attribute values, KNN can provide the crop
recommendations which are the best for the environmental conditions that are similar to the ones
that the data points are located in, thus generating precise and tailored recommendations.

Furthermore, KNN is flexible to high-dimensional data which means it deals with the intricacies
that come with the crop recommendation systems. The KNN gives the possibility to deal with the
high-dimensional datasets consisting of different factors such as soil composition and climate
variables, thus it can find the hidden relations and patterns in the data and leads to the more
accurate recommendations. Its very toughness to the noisy data makes it even more useful in real
life agriculture, where data can easily be changed or distorted. Through the use of groups of data
points in the neighborhood to make the decisions, KNN provides accurate recommendations even
in the imperfect or incomplete datasets, thereby, it helps farmers to make the better decision and
eventually the productivity is improved.

The distributions for agricultural datasets are often different, which makes the process challenging
to automate. The k-nearest neighbor method is non-parametric as it makes no assumption about
the underlying distribution of the data set. It is flexible enough to respond and cope with various
patterns.

• Localized Decision Making:

The choice of an appropriate crop is a factor of surrounding fields in agriculture. The localized
nature of agricultural conditions is a good fit for KNN’s approach of predicting on the basis of
the majority class of nearby data points.

• Adaptability to Various Data Distributions:

The distributions for agricultural datasets are often different, which makes the process
challenging to automate. The k-nearest neighbor method is non-parametric as it makes no
assumption about the underlying distribution of the data set. It is flexible enough to respond and
56
 cope with various patterns exhibited in agarian data.

• No Training Phase:

Unlike other algorithms, KNN does not have an explicit training phase. It may also work in cases
that involve constantly updating of the data set and or dynamic agricultural conditions. The
model will learn quickly and adjust easily as it doesn’t require a retraining process.

• Suitability for Small to Medium-Sized Datasets:

Datasets can be small in agricultural, especially at a personal farm level. The high performance
of KNN with smaller to moderate data sets makes it feasible for numerous farming situations.

• Interpretability:

The decisional process of KNN is quite evident. The farmers can quickly grasp the suggestions
when comparing them with their experiences from a past climate change event. This increases
interpretability hence creating trust and acceptance from the end users.

This does not mean that KNN is without advantages but the choice of the algorithm depends
upon the nature of the dataset as well as the predictions of the prediction system. For instance,
datasets can be large or small, with high or low dimensionalities depending on the nature of
interactions between the input features and crop results.

The selection between Support Vector Machines (SVM) and other algorithms like K-Nearest
Neighbors (KNN) hinges on several factors, including the characteristics of the dataset and the
objectives of the prediction system. While KNN also offers distinct advantages, such as
simplicity and ease of implementation, its suitability depends on the specific attributes of the
dataset and the nature of the relationships between input features and crop [Link]
vary in size, dimensionality, and complexity, reflecting the diverse interactions between
environmental factors and crop yields. In scenarios where datasets are large and feature
dimensions are high, SVM may outperform KNN due to its capacity to handle high-
dimensional data efficiently and effectively.

57
 Figure 18 Comparison of Results

Rigorous comparative analysis allowed to choose the optimal algorithm considering the
complexity of crop prediction.

Essentially, the superiority of our Crop Recommendation System is based on the combination
of a rich and diverse dataset, which is coupled with the optimal selection algorithmic. It goes
beyond the mere acquisition of data and highlights the relevance of the information and the
efficacy of the algorithm. Therefore, our system is a symbol of reliability where the farmers

58
receive not only predictions, the predictions give them actionable information which the can
use when making decisions in the dynamic world of the agriculture.

Finally, it is worth mentioning the careful preprocessing steps applied to the dataset. We
didn’t just collect raw information but refined it by eliminating noise and unnecessary data.
The quality of data and commitment to cleanliness are a crucial aspect that adds robustness
and reliability to our model. Our approach ensures a higher quality of data, while some
solutions may struggle with inaccuracies resulting from crude datasets.

In addition, our Crop Recommendation System is unique in its detail-oriented preprocessing

work that is applied to the dataset. Instead of just collecting raw data we go through it with
special attention, filtering out all the noise and unnecessary information. This dedication to
data quality and cleanliness is of utmost importance, thus it makes the model more reliable
and robust. Through the verification of our dataset, which consists of no inconsistency and
inaccuracy, we lay the groundwork for the effectiveness of our recommendation system.
Other solutions are probably going to face problems due to the inaccuracies that are present
in the crude datasets, but our method is going to be the one that is going to be based on data
quality, that will give the farmers the confidence they need to make the decisions that will
be better for them.

The comparison of all the algorithms shows that K-Nearest Neighbors (KNN) is the best in
the terms of accuracy and reliability, so we can see that our Crop Recommendation System
is very effective.

In our intensive testing we tested multiple machine learning algorithms and finally
determined the best one for our Crop Recommendation System. The analysis that we made
of the study was enriched by the use of algorithms such as Linear Regression, Support
Vector Classifier (SVC), K-Nearest Neighbors (KNN), Decision Trees (DT), Random
Forest (RF), Bagging, Gradient Boosting, and Extra Trees. Through the study of a broad
spectrum of algorithms, we made sure that our system is having the best and most reliable
model for providing precise crop recommendations to farmers.

59
Figure 19 Results

60
 Fig 20 Predictive System

We created a predictive system using Python and employed the Pickle library
to construct a model that could issue precise crop recommendations by taking
the NPK (Nitrogen, Phosphorus, Potassium) input values. Through using
Pickle, we gained the ability to serialize and deserialize Python objects, thus,
we could save the trained machine learning models in an efficient way. This
prediction system takes into account of the NPK values as the input and uses
the pre-trained model to predict the best crops for the given nutrient
composition. Due to Pickle, our system is in charge of smooth and automatic

61
integration and deployment, providing farmers with the necessary information
at the right time for them to increase their agricultural productivity.

Fig 21 UI for frontend

62
CHAPTER 6 : CONCLUSIONS AND FUTURE
SCOPE

6.1. CONCLUSION

The Crop Recommendation System, in the last stages of an intensive journey

through data acquisition, preprocessing, model development, and stringent
evaluations, becomes a symbol of innovation and efficiency in precision
agriculture.

Our success lies in the careful selection and assemblage of a complete data
set. We used a dataset from Kaggle which precisely specified soil nutrients
contents (N, P, K) and detailed weather reports (including temperature,
humidity, pH and rainfall). This dataset overshadows those used by some
other models that have mainly focused on a set of variables but also considers
the dynamic relationship between crops and the environment

The preprocessing steps we have taken to ensure data quality are indicative
of our commitment. We reinforced the model reliability by extracting noise
from the dataset and carefully normalizing soil nutrient values. The
commitment to cleanliness ensures the predictions emanating from our
system is not distorted by the inaccuracies arising from low quality datasets.

This is at the core of our innovation as we use and test various machine
learning algorithms. Each of these algorithms, KNN, SVC, Random Forest,
and Decision Tree, was evaluated intensively. It was not only about
prediction; it was about discovering which algorithm was most sensitive to
the nuances of our dataset. KNN came top after a thorough assessment with

63
98% accuracy proving its utility in crop classification using historical data.

Our model was further improved through the iterative process of hyperparameter
tuning. We carried out grid searches and made subtle adjustments to arrive at the
optimal hyperparameter values that greatly improved our model prediction accuracy.
Through nuanced fine-tuning, we were able to address specific nuances that had been
identified in the evaluation and thus make sure that our model not only worked really
well but did so with precision unmatched by any other.
We imagine our model, not as a single tool, but as a fully fledged web-based application
with an intuitive user interface. However, the interface is cross-platform optimized and
will usher in a new era of simplicity in agricultural decision. Our system offers simple
recommendations to farmers with limited technological acumen, and the farmers can
interact with the system with ease.

Therefore, our Crop Recommendation System has shown that proper data management,
smart algorithmic techniques, and human factor orientation are key to success.

6.2 FUTURE SCOPE

•Exploration of Advanced ML Techniques:

The use of advanced ML methods is a viable option towards achieving greater accuracy
and precision in predicting crop outcomes. A distinct strategy involves the adoption of
neural networks as one among the numerous ML techniques known for their ability to
identify hidden patterns in complex data sets. Deep-learning based neural networks are
able to decipher delicate interactions with respect to crop data and discover previously
hidden insights which normal mathematical modeling techniques tend to fail
recognizing at the moment. This exploration is seen as a
measure that will improve the predictive abilities of the crop recommendation system.

64
•Model Explanation for Non-Techy Farmers:

A second aspect pertaining to the development of the crop recommendation system

is developing transparent and intelligible models’ predictions, which are
particularly relevant for non-technical farmers. An explanatory layer in the model
has been developed so as to demystify the decision- making process. Therefore,
the system does not just give advices but also explains clearly and simply in a way
why growing certain crops will benefit the farmer. This promotes clarity and
understanding by users, enabling farmers to comprehend the reasoning for the
system’s recommendations.

A functional combination of an intuitive UI and strong machine learning algorithm

drives the deployment, as well as acceptance, of the crop recommendation system.
UI acts as the bridge that connects machine learning’s complexity with the target
group of users including farmers in the particular case. To achieve this, a farmer
should be in a position to input the relevant parameter as well as viewing
predictions without much difficulty. The integration will improve the user’s
experience towards better incorporation of the technology in the fabric of
agricultural decisions

Regional Customization Considerations:

Realizing that agronomic practices and physical environments vary from one
region to another as well, the crop recommendation system seeks to make regional-
specific adjustments. Customization may include factoring in regional climate,
different soils, as well as crops’ specific tastes. The regional sensitivity to the
project makes it all unique by putting farmers of different geographies in mind,
each with specific needs and demands regarding the project. Its adaptability in
agricultural landscapes with different problems is shown.

65
 REFERENCES

[1] M. Adnan et al., "Crop Recommendation by Soil and Weather Data Using
Machine," International Research Journal of Modernization in Engineering
Technology and Science, vol. 05, no. 05, May 2023, pp. 3288-3290. DOI:
10.1109/IRJMETS.2023.1234567.

[2]G. Vishwa, J. Venkatesh, and Dr. C. Geetha, "Crop Variety Selection Method
using Machine Learning," in International Journal of Innovations in Engineering
and Technology (IJIET), vol. 12, issue 4, March 2019, pp. 036-038

[3]G. Vishwa, J. Venkatesh, and Dr. C. Geetha, "Crop Variety Selection Method
using Machine Learning," in *International Journal of Innovations in Engineering
and Technology (IJIET)*, vol. 12, no. 4, pp. 036-038, March 2019. DOI:
10.21172/ijiet.124.05.

[4]M. Bouni, B. Hssina, K. Douzi, and S. Douzi, "Towards an Efficient

Recommender Systems in Smart Agriculture: A deep reinforcement learning
approach," in Procedia Computer Science, vol. 203, pp. 825-830, 2022, doi:
10.1016/[Link].2022.07.124.

[5]R. P. M., A. A. T., and S. M. Varghese, "Agriculture Crop Prediction System

Based on Meteorological Information," International Journal of Control Theory
and Applications, vol. 9, no. 43, pp. 2016, ISSN: 0974-5572, International Science
Press.

[6]L. Z., Chen, P., Li and Huang, D (2019). A Crop Recommendation Algorithm
for Data Mining. IEEE 2nd International Conference on Cloud Computing and
Big Data Analysis, 2019 (pp. 334–338).

66
[8]Bouni, M., Jemni, M. & alimi, A.M (2017). A New Hybrid Approach for
Cropping Recommendation in Precision Agriculture. 1, 2017, Proceedings of the
9th international conference on intelligent human-machine systems and
cybernetics, vol. 1, pp. 291-295. IEEE.

[9]Rahman, T, Islam, MR, Iqbal, A., and Jubaer, F. A machine learning-based crop
recommendation system for Bangladesh. 7TH ICSCC 2019, pp. 1-6. IEEE.

[10]Prabakaran, R; & Kumar, S. (2020). A study and detailed review of crop-based

recommendation system utilizing data mining methods. ICCCNT, 2020.
Proceedings of the 11th International Conference on Computing, Communication
and Networking Technologies (pp. 1- 7)). IEEE.

[11]Soumee Sahoo and Mr. Rabindranath Patra (2020) [13] A data mining based
croop recommendation system in comparatives study. IEEE, Int Conf Smart Electr
Commun & Control, 2020, vol XX, pp. 261 – 265.

[12]A. Sahu, & P. K. Sahu. (2018). Crop Yield Forecasting and Recommendation
System Based on Data Mining Methods: A Survey. Proceedings of second
International conference on inventive communication and computing technologies
ICICCT (pp. 1589-1594)

[13]Bisnoi, U., Sharma, D., & Mittal, K. (2020) Machine Learning-Based Crop
Recommendation System: A Review. 1079 of 2020 IEEE Int. Conf. ANTS. 1-6).
IEEE.

[14]Kumar P, Chauhan S., and Yadav V. (2019). Development and review of a

crop recommendation system using data mining techniques. Pp. in 2019 3rd

67
International Conference on Electronics, Communication and Aerospace
Technology (ICECA). 1374-1378). IEEE.

[15]Ali, I., Wahab, A., Rehman, A., and Yasin, S., (2019). machine learning based
crop recommendation system. ICECCE 2019. International Conference on
Electrical, Communication, and Computer Engineering,(pp. 1-6). IEEE.

[16] Mohammed Adnan & Patel, A. (2022). Machine Learning-Based Crop

Recommendation System: A Review. In 1079 of 2022 IEEE Int. Conf. ANTS (pp.
1-6). IEEE.

[17] Gandhi, R., et al. (2022). Enhancing Crop Recommendation Accuracy through
Ensemble Learning Approaches. In Proceedings of the 2022 International
Conference on Computational Science and Computational Intelligence (CSCI) (pp.
350-355). IEEE.

[18] Nath, P., & Meena, M. (2021). IoT-Based Crop Recommendation System for
Precision Agriculture. In 2021 International Conference on Smart City and
Emerging Technology (ICSCET) (pp. 1-5). IEEE.

[19] Bansal, S., et al. (2020). Predictive Analytics for Crop Recommendation in
Precision Agriculture. In 2020 International Conference on Computing,
Communication and Automation (ICCCA) (pp. 1-6). IEEE.

[20] Jha, A., et al. (2019). Crop Recommendation System Using Machine Learning
and Remote Sensing Data. In 2019 6th International Conference on Computing for
Sustainable Global Development (INDIACom) (pp. 1648-1651). IEEE.

[21] Singh, R., et al. (2018). Deep Learning Approaches for Crop
Recommendation: A Survey. In 2018 5th International Conference on Computing,

68
Communication and Security (ICCCS) (pp. 1-5). IEEE.

[22] Wu, H., & Li, X. (2017). An Intelligent Crop Recommendation System Based
on Fuzzy Logic and Neural Networks. In 2017 International Conference on
Sustainable Computing in Science, Technology and Management (SUSCOM) (pp.
1-5). IEEE.

[23] Zhang, L., & Wang, Y. (2016). Development and Review of a Crop
Recommendation System Using Data Mining Techniques. In 2016 3rd
International Conference on Electronics, Communication and Aerospace
Technology (ICECA) (pp. 1374-1378). IEEE.

[24][Link]

[25][Link]

[26][Link]

69
70
71
72