Advanced Algorithms in Al & ML Course Code: 22683 Program Name —_: Diploma in Artificial Intelligence and Machine Learning Program Code: AN Semester : Sixth Course Title : Advanced Algorithms in AI & ML Course Code 222683 1. RATIONALE Machine Learning is a branch of Computer Science that uses algorithms to imitate the way in which humans learn. It uses statistical methods to train algorithms and make predictions. Machine learning is one of the most in-demand Data Science skills, which allows data scientists to increase the accuracy of predictions of software applications, without explicitly programming them to do so. These algorithms make use of historical data to predict output values and these insights and predictions enable businesses to make smart decisions 2. COMPETENCY The aim of this course is to help the student to attain the following industry identified competency through various teaching learning experiences: © Implement AI-ML algorithm using Python, 3. COURSE OUTCOMES (COs) The theory, practical experiences and relevant soft skills associated with this course are to be taught and implemented, so that the student demonstrates the following industry oriented COs associated with the above mentioned competency: Implement preprocessing steps on data to make it ready for analysis, Implement Naive Bayes and Random Forest algorithm. Describe Support Vector Machines and K Nearest Neighbors. Apply clustering algorithm and Dimensionality Reduction, Describe ANN and basic hyper parameters of Deep Learning. Classify Sequential and Image Data of Deep Learning. pepoge TEACHING AND EXAMINATION SCHEME, Teaching Examination Scheme Scheme (In Hours) Theory Practical Credits | Paper |__ESE PA Total ESE PA Total ae (L+T+P)| Hrs. Max |Min |Max |Min} Max |Min|Max | Min |Max | Min | Max | Min 3/0} 2 5 70 | 28 | 30*] 0 | 100 | 40|25@| 10 | 25 | 10] so | 20 (**) marks should be awarded on the basis of internal end semester theory exam of 50 marks based on the specification table given in S. No. 9. (~): For the practical only courses, the PA has two components under practical marks ie. the assessment of practicals (seen in section 6) has a weightage of 60% (i.e.30 marks) and micro- ‘project assessment (seen in section 12) has a weightage of 40% (i.e.20 marks). This is designed to facilitate attainment of COs holistically, as there is no theory ESE. MSBTE ~ Final Copy Dt. 11.07.2023 Page lof 8Advanced Algorithms in AL& ML Course Code: 22683 COURSE MAP (with sample COs, Learning Outcomes i.e. LOs and topics) This course map illustrates an overview of the flow and linkages of the topics at various levels, of outcomes (details in subsequent sections) to be attained by the student by the end of the course, in all domains of learning in terms of the industry/employer identified competency depicted at the centre of this map. CSE. Cm aia ey swag, cater Dearbe ANN NO eS. ie Se ac_ apcay eee _ Cease S) ae " eae ae, ae Geist) Seatit 7 ———— ey Nhetigen “elena, Figure 1 - Course Map 6. SUGGESTED PRACTICALS/ EXERCISES The practicals/exercises/tutorials in this section are psychomotor domain LOs (ie. sub- components of the COs) are to be developed and assessed in the student to lead to the attainment of the competency. sr. Practical Exercises Unit Eee No. (Learning Outcomes to be achieved through practicals) No. | penuied a. Installation of Python (PyCharm) | b. Installation of Python scikit lean for ML Pa 1 |e. Installation of Tools and Libraries ( Jupyter I Sm Notebook/Matplotlib/Numpy/Pandas) d._Use of google colab (hitps://colab.research google.com’) MSBTE — Final Copy Dt. 11.07.2023 Page 2 of 8Advanced Algorithms in Al & ML. Course Code: 22683 [ sr. Practical Exer Unit | om | No. | (Learning Outcomes to be achieved through practicals) No. Require i 2 _| Study Different datasets such as Iris dataset ,Titanie dataset Imdb ; Movies dataset - Perform following operations :( Assume suitable data/dataset if needed). a. Write program to read dataset ( Text,CSV,ISON.XML) 3 |b. Which of the attributes are numeric and which are categorical? | aa c. Performing Data Cleaning, Handling Missing Data, Removing Null data dd. Rescaling Data v. Encoding Data ce. Feature Selection 1 _ | Write a python Programming code to apply Naive Bayesian . aa algorithm for classification using suitable data/dataset 5 _| Write a python Programming code to implement decision tree for [47 | gps classification using suitable data/dataset.. Implement the Random Forest Algorithm using following Steps a. Data Preprocessing Step | b. Litting the Random Lorest Algorithm to the Liamung Set | 6 | c. Predicting the Test Set Result 0 04 d. Creating the confusion Matrix e. Visualizing the training set result | £. Visualizing the test set result 7 _ | Implement unsupervised machine Teaming algorithm (KNN) in python | 77 a on dataset to cluster data. (Assume suitable dataset) §g _ | Implement unsupervised machine leaming algorithm (Clustering—K | ry | ge Means) in python on dataset to cluster data. (Assume suitable dataset). 9_| ImplemenvSimulate Back propagation/feed forward neural network. [Vv | od 10 | Study of Classification of Dog images and Cat images categories = aa using suitable dataset (foe ex imagenet dataset) Total 32 ii, “*: compulsory practicals to be performed. Note i. Given in above tables is suggestive list of practical exercises. Teachers can design other similar exercises, work should be done as per suggested sample below: Assessment of the ‘Process’ and ‘Product’ related skills in the laboratory/workshop/field S.No. | Performance Indicators Weightage in % 1__| Import packages and Libraries of Python. 20 2 | Use Python to create, edit, assemble and link the programs. 40 3__| Debug, test and execute the programs 20 4__| Able to answer oral questions. 10 5__[ Submission of report in time. 10 Total | 100-1 SA MSBTE — Final Copy Dt. 11.07.2023 Page 3 of 8Advanced Algorithms in Al & ML Course Code: 22683, Additionally, the following affective domain LOs (social skills/attitudes), are also important constituents of the competency which can be best developed through the above mentioned laboratory/field based experiences: Handle command prompt environment. Experiment with Python Plan, develop, assemble, link, debug and test the programs. Demonstrate working as a leader/a team member. Maintain tools and equipment. Follow ethical practices. The development of the attitude related LOs of Krathwohl’s ‘Affective Domain Taxonomy’, the achievement level may reach: © ‘Valuing Level’ in 1" year © ‘Organizing Level’ in 2™ year and © ‘Characterizing Level’ in 3" year. pepege 7. MAJOR EQUIPMENT/ INSTRUMENTS REQUIRED The major equipment with broad specification mentioned here will usher in uniformity in conduct of experiments, as well as aid to procure equipment by authorities concerned. S Equipment Name with Broad Specifications PrO. Sr. No. 1 | Hardware: Personal computer, (3 preferable), RAM minimum 4 GB onwards. Forall 2 | Operating system: Windows 7 onward Experiments 3 | Libraries of Python 8. UNDERPINNING THEORY COMPONENTS The following topics/subtopics should be taught and assessed in order to develop LOs in cognitive domain for achieving the COs to attain the identified competency. Unit Major Learning Outcomes Topies and Sub-topies (in cognitive domain) Unit 1 Ta, Select a suitable model for _| 1-1. Introduction: Selecting a model Model the given data with 1.2. Training a model for supervised selection and | justification. learning Features ~ understand your feature 1b Describe the process of data better, Feature extraction and engineering using supervised learning engineering, on the given data. 1.3. Feature engineering on ~ numerical le. Describe the process of data & categorical data & text data feature extraction and 1.4, Feature scaling, Feature selection engineering on the given data. 1d. Compare Feature Engineering for the given type of data. le. Select feature scaling, feature selection, dimensionality reduction in the given situation with justification. ‘Unit 2a. Classify the yiveu data 2.1 Naive Bayes’ [xf Are lm] MSBTE — Final Copy Dt. 11.07.2023 Page 4 of 8 aN Kea Sid os Ove Ay LiAdvanced Algorithms in Al & ML Course Code: 22683 Naive Bayes, Decision Tree Major Learning Outcomes __Gin cognitive domain) using Bayesian method with stepwise justification. | 2b. Describe Working of Decision tree algorithm 2c. Enlist application of Random Forest Algorithm. Topics and Sub-topics Bayes Theorem. of Naive Bayes Bayes classifier, Applying Bayes Theorem, Advantages and Disadvantages of Naive Bayes classifier, Application of Naive Bayes, Implementation of Naive Bayes classifier Working 2.2 Decision Tree Decision Tree Diagram, Why Used decision tree? Working of Decision tree algorithm, attributes sclection Measures (ASM), Advantages and Disadvantages of Decision tree, Implementation of Decision ‘Tree 2.3 Random Forest Why use Random Forest? Working of Random Forest Algorithm, Application of Random Forest Algorithm, Advantages and Disadvantages of Random Forest algorithm, Implementation of Random Forest algorithm Unit-II Supervised Learning: Support Vector Machines, K Nearest Neighbors 3a, Describe Support Vector Machines. 3b. Enlist advantages and disadvantages of KNN algorithm, 3.1 Support Vector Machines: Types of SVM, How does SVM work? Advantages and Disadvantages of Decision tree, Implementation of SVM 3.2K Nearest, Neighbors: Need of KNN algorithm, Working of KNN Algorithm, Advantages and Disadvantages of KNN algorithm, Implementation algorithm, of KNN Unit-1V Unsupervised learning: Clustering Algorithms 4a, Describe the performance analysis of clustering for the given situation, 4b. Describe Dimensionality Reduction 4.1 K-Means Clustering: What is K-means Clustering? Working of _-~—-K-means Algorithm, Failure of K-means a] Implementation off MSBTE — Final Copy Dt. 11.07.2023 Page 5 of 8Advanced Algorithms in AL & ML Course Code: 22683 Unit _(in cognitive domain) Major Learning Outcomes Topics and Sub-topics algorithm 4.2 Dimensionality Reduction: # Introduction to Dimensionality Re- duction, Subset Selection, © Introduction to Principal Component Analysis. Unit V Introduction to Deep Learning 5a. Describe ANN Concepts 5b. Leam Hyper parameter basics. 3:1 Introduction Artificial Neural Network, ‘* Perceptron EX-OR problem, * Feed Forward and Back Propagation, Losses * Activation Funetion, GPU Training asics Hyper parameter * Selecting number of Neurons, + Activation Functions, © Layers using Greedy Search and Random Access Unit VI Deep Learning for Sequential and Image Data 6a, Implement Deep Learning for sequential Data 6b. Implement Deep Learning for Image Data 6.1Sequential Data: + RNN, LSTM, LSTM-GRU, ‘© Introduction to Transformers, GPT | 6.2 Image Data : * CNN, (Resnet , VGG) Pre-trained Neural Networks, Transfer Learning, «Fine Tunin; Note: To attain the COs and competency, above listed Learning Outcomes (LOs) need to be undertaken to achieve the ‘Application Level’ of Bloom’s ‘Cognitive Domain Taxonomy’: 9, _ SUGGESTED SPECIFICATION TABLE FOR QUESTION PAPER DESIGN ‘ __, | Distribution of Theory Marks x Unit Title ae a | 8 3 Level | Level | Level | Marks 1 _ | Model selection and feature 06 aula leetce engineerin 11_| Supervised Teaming: Naive Bayes, os wale lecmlrn Decision Tree Supervised Leaming: Support Vector Il | Machines, K Nearest Neighbors wy a a 2 a ry _| Unsupervised learning: Clustering rs aaa ele Algorithms V_| introduction to Deep Learning 08 2 [4/6] 2 ‘yy_| Deep Leaming for Sequential and i awleelnamles Image Data Total! 48 | 16 | 26 Legends: R=Remember, U=Understand, A=Apply and above (Bloom's Revise MSBTE ~ Final Copy Dt. 11.07.2023 Page 6 of 8Advanced Algorithms in AL & ML Course Code: 1683, Note: This specification table provides general guidelines to assist student for their learning and to teachers to teach and assess students with respect to attainment of LOs. The actual distribution of marks at different taxonomy levels (of R. U and A) in the question paper may vary from above table. This specification table also provides a general guideline for teachers 1o frame internal end semester practical theory exam paper which students have to undertake, 10. SUGGESTED STUDENT ACTIVITIES Other than the classroom and laboratory learning, following are the suggested student-related co-curricular activities which can be undertaken to accelerate the attainment of the various outcomes in this course: a. Prepare journals based on praotioal porformed in laboratory. , Library/E-Book survey regarding Advanced Algorithm in AI & ML used in Computer industries. ¢. Prepare power point presentation for showing different types of Advanced Algorithm in Al & ML Applications, 11, SUGGESTED SPECIAL INSTRUCTIONAL STRATEGIES (if any) ‘These are sample strategies, which the teacher can use to accelerate the attainment of the various outcomes in this course: a. Massive open online courses (MOOCs) may be used to teach various topies/sub topics. b. ‘Lin item No. 4 does not mean only the traditional lecture method, but different types of teaching methods and media that are to be employed to develop the outcom: c. About 15-20% of the topics/sub-topics which is relatively simpler or descriptive in nature is to be given to the students for self-directed learning and assess the development of the LOs/COs through classroom presentations (see implementation guideline for details). 4. With respect to item No.10, teachers need to ensure to create opportunities and provisions for co-curricular activities. €. Guide student(s) in undertaking micro-projects. £. No. of practical’s selection to be performed should cover all units. 12, SUGGESTED MICRO-PROJECTS Only one micro-project is planned to be undertaken by a student assigned to hinv/her in the beginning of the semester. S/he ought to submit it by the end of the semester to develop the industry oriented COs. Each micro-project should encompass two or more COs which are in fact, an integration of practicals, cognitive domain and affective domain LOs. The micro- project could be industry application based, internet-based, workshop-based, laboratory-based or field-based. Each student will have to maintain dated work diary consisting of individual contribution in the project work and give a seminar presentation of it before submission, The total duration of the micro-project should not be less than 16 (sixteen) student engagement hours during the course. In the first four semesters, the micro-project could be group-based. However, in higher semesters, it should be individually undertaken to build up the skill and confidence in every student to become problem solver so that s/he contributes to the projects of the industry. A suggestive list is given here. Similar micro-projects could be added by the concerned faculty: a. Credit Card Fraud Detection Using Classification Algorithm b. Anomaly detection Using KNN Machine learning c. Image Classification using Support Vector machine d. Classification based on Decision Tree Application of SVM algorithm for Cancer detection MSBTE — Final Copy Dt. 11.07.2023 Page 7 of 8Advanced Algorithms in Al & ML. Course Code: 22683 13.__ SUGGESTED LEARNING RESOURCES. Title of Book Author Publication , Introduction to Machine ‘Andreas C. Miller & | O°Reilly Media, Inc | Learning with Python Sarah Guido | DreamTech, First Edition, 2 | Machine Learning in Action | Peter Harrington, 2012 | 13: 978- 1617290183 3 | Machine Leaning ‘Tom M Mitchell, erate at inaaiT ) Machine Leaning Step-by- 4] Ng nei ee Rudolph Russell | CreateSpace Independent with Python Mastering Machine Learning with Python in Six Steps A Apress Publication 5 . Manohar 5 | Practical Implementation epee ea ISBN-13 (pbk): 978-1- Guide to Predictive Data nas 4842-2865-4 Analytics Using Python Practical Machine Learning ‘Apress publication with Python A Problem- Dipanjan Sarkar, ISBN-13 (pbk): 978-1- 6 | Solver’s Guide to Building | Raghav Bali, Tushar | 4842-3206-4 ISBN-13 Real-World Intelligent Sharma (clectronic): 978-1-4842- Systems 3207-1 7 | Machine Learning using Manaramjan Pradhan, | Wiley India Python U Dinesh Kumar ISBN: 978-81-265-7990-7 ean FRANCOIS Manning Publications 8 | Deep Leaming withPython | Con pT ISBN 9781617294433 9 | Python Deep Learning Second | Ivan Vasilev, Daniel | Packt Publishing Edition Slater ISBN 978-1-78934-846-0 14. SOFTWARE/LEARNING WEBSITES a. _https:/vww.pdfétive.com/machine-leamning-for-absolute-beginners-e188007429 html b. hups:/vww.geeksforgeeks.org/ml-fuzzy-clustering/ c._https:/vww.pdfadrive.com/machine-leaming-step-by-step-guide-to-implement- machine-leaming-algorithms-with-python-d158324853.html 4._ https:/www.pdfédive.com/machine-leaming-for-absolute-beginners- 188007429. html e._https://www.geeksforgeeks.org/ml-fuzzy-clustering/ £_https://www.pdférive.com/machine-learning-step-by-step-guide-to-implement- machine-learning-algorithms-with-python-d158324853.html g._https://machinelearningmastery.com/classification-as-conditional-probability-and-the- naive-bayes-algorithm/ (Practical) h,_https://www.geeksforgecks.org/naive-bayes-classifiers/(Practical) i. _htaps://www javatpoint.com/machine-learning-naive-bayes-classifier (Practical) j.https:/Avww.javatpoint.com/machine-leaming-random-forest-algorithm (Random Forest) k MSBTE — Final Copy Dt. 11.07.2023, https://leaming.oreilly.comv/library/view/deep-learning- wih TB 129686 6 FT hitps://vww.deeplearningbook.org/