ROAD ACCIDENT ANALYSIS
USING MACHINE LEARNING
Presented by:
Maneesha Kumari
Roll no :-2104310105006
Under the supervision of:
Mr. Vivek Rai
Dept. of computer science and engineering
B.N. COLLEGE OF ENGINEERING &TECHNOLOGY,LUCKNOW
Affiliated to
Dr. A.P.J. ABDUL KALAM TECHNICAL UNIVERSITY,LUCKNOW
CONTENT
Area of research
Introduction
Objective
Existing system vs proposed system
Literature Review
Material and methods used
Issues and challenges
Result
Conclusion
References
Area of research
Road Accident analysis and
prevention
Introduction
• One of the most complicated and difficult daily needs are overland transportation. In
India, more than 150,000 people are killed each year in traffic accidents.
• That’s
about 400 fatalities a day and far higher than developed auto markets like the US,
which in 2016 logged about 40,000.
• Everyyear over 1 million vehicles are added to traffic on an average 1.2 million People
have died and over 50 million people have been injured in road accidents in the world
every year.
• Studies on traffic have executed that road accidents and death- laceration ratio will
increase. Design and control of traffic by advanced systems come in view as the
important need.
• Assumption on the risks in traffic and the regulations and interventions in the end of
these assumptions will reduce the road accidents.
OBJECTIVE
The main objective of the road accident prediction system:
• Analyze the previously occurred accidents in the locality which will help us
to determine the most accident-prone area and help us to set up the
immediate required help for them.
• To make predictions based on constraints like weather, pollution, road
structure, etc.
Existing system vs proposed
system
Existing Proposed
system system
The current system provides little information In proposed system accident data records
regarding the number of injuries and accidents are used to build models, which can help us
The current system only partially provides the understand things like drivers' behavior,
information with the desired characteristics. road conditions, lighting, weather, and other
Both quantity and quality are lacking in the factors. This may assist users in calculating
collected data. the safety measures that are useful for
preventing accidents.
There is no specific form for gathering
information about road accidents. At all levels, Predictions can be made thanks to this
narrative reports are typically used to report analysis, which aims to highlight the most
the accident important data in a road traffic accident.
Literature
Review
• Ossenbruggen and others [ With the intention of using these models to carry
out a risk assessment of a specific region, 24] utilized a logistic regression
model to identify statistically significant factors that predict the probabilities
of crashes and injury crashes. Their research demonstrated that village sites
are safer than residential or commercial ones.
• Abdelwahab and others analyzed the accident data for the Central Florida
region from 1997 [2]. The focus of the analysis was on car accidents that
took place at intersections with signals. There were three levels of severity
for the injuries: no injury, a possible injury, and a disability. They found that
Multi-layered Perceptron (MLP) classification accuracy is higher than that of
Fuzzy ARTMAP when they compared their results.
• Bedard and co. utilized multivariate logistic regression to identify the
independent contribution of driver, crash, and vehicle characteristics to the
fatality risk of drivers [3]. They discovered that reducing speed, reducing the
number and severity of driver-side impacts, and increasing seatbelt use may
reduce fatalities. To ascertain the connection between accident notification
times and fatalities, Evanco carried out a multivariate population-based
statistical analysis [6].
• The study found that the length of time it takes to notify drivers of an
accident is a significant factor in the number of fatalities resulting from
collisions on rural roads. Ossiander and others utilized Poisson regression to
examine the relationship between the speed limit increase and the fatal crash
rate (fatal crashes per vehicle mile traveled) [13]. In Washington State, they
discovered that an increase in the speed limit was linked to a higher rate of
fatal crashes and an increase in fatalities on freeways.
Material and methods
used
Technology used:-
• Python
• Django
• MySql
• WampServer 2.4
Algorithm used:-
K-means clustering
Algorithmic steps for k-means clustering
LetcX = {x1,x2,x3,……..,xn} be thecset of data pointscand V = {v1,v2,…….,vc}
be thecset of centers.
1) Randomlycselect ‘c’ clusterccenters.
2) Calculate thecdistance between eachcdata point and clusterccenters.
3) Assign thecdata point to the clusterccenter whose distancecfrom the cluster
center iscminimum of all the clusterccenters..
4) Recalculatecthe new clusterccentercusing:
where, ‘ci’ representscthe number of datacpoints in ith cluster
5) Recalculatecthe distance betweenceach data point and newcobtained
cluster centers.
6) If no datacpoint was reassignedcthen stop, otherwisecrepeat from step
System Design
• ARCHITECTURE DIAGRAM
• Component diagram
Issues and
challenges
proper selection of features which can cause a road accident
Removal of unwanted features
Collected data should be valid and related to real life accident
situation
Removal of outlier instances
Result
The dataset used in the project to predict road accidents is based on values. Since
there are a lot of columns and rows in this dataset, the forward fill method and the
classification algorithm will be used to fill in all of the null values. The k-means
clustering algorithm will be used in this classification algorithm.
The final outcome of the Road Accident Prediction is the percentage of accidents in
a specific area. The algorithm converges faster and is more accurate when there are
fewer features. The final outcome of the Road Accident Prediction is the percentage
of accidents in a specific area. The least amount of error is then obtained by
applying k-means clustering to these features.
Data set( cont..)
Vehicle analysis graph (cont..)
Speed limit analysis graph (cont..)
Day of week analysis graph
K - means clustering
graph( cont..)
Conclusion
The arrangement exactness acquired in our analyses uncovers that, for
the non-debilitating injury, the weakening injury, and the lethal injury
classes, the mixture approach performed better compared to brain
organization, choice trees and backing vector machines. The hybrid
approach outperformed the neural network in both the no injury and
possible injury classes. K means clustering might be the most effective
method for directly modeling the classes of no injury and possible
injury.
References
• [1] Abdel-Aty, M., and Abdelwahab, H., Analysis and Prediction of Traffic Fatalities Resulting From
Angle Collisions Including the Effect of Vehicles’ Configuration and Compatibility. Accident Analysis
and Prevention, 2003.
• [2] Abdelwahab, H. T. and Abdel-Aty, M. A., Development of Artificial Neural Network Models to
Predict Driver Injury Severity in Traffic Accidents at Signalized Intersections. Transportation
Research Record 1746, Paper No. 01-2234.
• [3] Bedard, M., Guyatt, G. H., Stones, M. J., & Hireds, J. P., The Independent Contribution of Driver,
Crash, and Vehicle Characteristics to Driver Fatalities. Accident analysis and Prevention, Vol. 34, pp.
717-727, 2002.
[4] Buzeman, D. G., Viano, D. C., & Lovsund, P., Car Occupant Safety in Frontal Crashes: A Parameter Study of
Vehicle Mass, Impact Speed, and Inherent Vehicle Protection. Accident Analysis and Prevention, Vol. 30, No. 6, pp.
713-722, 1998.
[5] Dia, H., & Rose, G., Development and Evaluation of Neural Network Freeway Incident Detection Models Using
Field Data. Transportation Research C, Vol. 5, No. 5, 1997, pp. 313-331.
[6] Evanco, W. M., The Potential Impact of Rural Mayday Systems on Vehicular Crash Fatalities. Accident Analysis
and Prevention, Vol. 31, 1999, pp. 455-462.
[7] Hand, D., Mannila, H., & Smyth, P., Principles of Data Mining. The MIT Press, 2001.
[8] Kim, K., Nitz, L., Richardson, J., & Li, L., Personal and Behavioral Predictors of Automobile Crash and Injury
Severity. Accident Analysis and Prevention,
Vol. 27, No. 4, 1995, pp. 469-481.
[9] Kweon, Y. J., & Kockelman, D. M., Overall Injury Risk to Different Drivers: Combining Exposure, Frequency, and
Severity Models. Accident Analysis and Prevention, Vol. 35, 2003, pp. 441-450.
