Road accident Prediction Using Data Mining Techniques

Anurag S Tippa [MY.SC.U3BCA21133]

Thilakesh A [MY.SC.U3BCA21106] Supervisor:Mrs.Keerthika.K Assistant Professor, Computer
Mahadev Prasad M [MY.SC.U3BCA21124]
Science Engineering, School of Arts and Science.
Gowtham M [MY.SC.U3BCA21104]

RESULTS
ABSTRACT INTRODUCTION
Road accidents are a global concern, causing loss of life, Decision tree: DISCUSSION
This study uses a variety of data sources, injuries, and economic costs. Data mining techniques, 1. The study aimed to predict if Karnataka districts would Random Forest outperformed Decision Tree with 91%
such as past accident reports, incorporating data analysis and machine learning, offer have over 750 total accidents using a decision tree accuracy compared to Decision Tree's 61%.
meteorological data, and traffic patterns, a powerful means to predict and prevent accidents. classifier. - Random Forest's superior performance demonstrates its
to anticipate road accidents using data This approach involves analyzing historical accident 2. Features included total accidents, fatalities, severe robust predictive capabilities.
mining techniques. data to identify patterns and risk factors, injuries, minor injuries, and total injuries, with 'Vehicle to - The findings highlight Random Forest as a preferred
The study uses machine learning methods In India, where road safety challenges are pronounced, Vehicle Total Accidents' as the target variable. choice for tasks needing high precision and accuracy.
for feature selection and thorough the adoption of data mining for accident prediction is on 3. The decision tree classified districts as 'False' for - The results underscore Decision Tree's limitations in
analysis, including decision trees and the rise. Government agencies, including the Ministry of accidents ≤750 and 'True' for accidents >750, visualized capturing complex patterns and nuances within the dataset.
neural networks. Road Transport and Highways, utilize predictive in tree.
To maintain data quality, the dataset is models to inform safety programs. Collaborative efforts 4. It listed districts with over 750 accidents, using this
meticulously preprocessed to handle involving universities, tech companies, and NGOs are threshold as a baseline for identifying patterns leading to CONCLUSIONS
outliers and missing values. actively contributing to road safety improvement. high accident rates.
The motivations in India stem from high accident rates, 5. The tree's visualization highlighted key features and Conclusion:
Complex interactions among contributing
substantial economic impact, limited resources, urban thresholds used for classification, aiding in understanding - The study uses Decision Tree and Random Forest algorithms to
components can be understood through
congestion, and the need for infrastructure factors contributing to higher accident rates. analyze road accident data.
the use of advanced analytics and
development. Predictive models aid in optimizing 6. This information helps in targeted interventions and - Random Forest achieves 91% accuracy, outperforming Decision
visualisation technologies.
resource allocation, supporting efficient traffic preventive measures. The Decision Tree method had an Tree.
Real-time and historical data are used to
management, informing infrastructure projects, and accuracy of 51%, the lowest in the experiment. - A comprehensive dataset from historical reports, meteorological
thoroughly evaluate and validate the
influencing policy changes. Overall, data-driven accident Random Forest:- data, and traffic patterns is utilized.
model's performance.
prediction contributes to public safety, reduces 1. Used attributes: "Vehicle to vehicle total injured" and - Advanced data mining techniques ensure thorough analysis and
By enabling proactive accident risk
economic burdens, and fosters responsible road use "Vehicle to animal persons minor injury." precise predictions.
minimization, the suggested predictive
through awareness campaigns. 2. Achieved a notable accuracy of 91%. - The model serves as an early warning system, aiding accident
model functions as an efficient early
3. Demonstrated high efficacy in analyzing and predicting prevention, policy-making, traffic management, and urban
warning system for law enforcement and
factors associated with road accidents. planning.
motorists.
4. Highlights Random Forest's robust performance and - Real-time and historical data integration enhances the model's
The study improves overall road safety by METHODS AND MATERIALS value for law enforcement and motorists.
assisting with policy formation, traffic suitability for complex predictive tasks in road safety
Decision tree: When proposing to use the ID3 algorithm analysis. - Data mining techniques significantly improve road safety and
management, and urban planning.
for traffic accident prediction, it is important to describe reduce societal and economic impacts.
The purpose of this study is to offer a
the specific objectives, methods and expected results of Future Scope:
useful tool for predicting and preventing
the study or project. Below is an example of a proposed - Further research can explore factors like vehicle interactions,
accidents, ultimately reducing their
action plan for using the ID3 algorithm in traffic injuries, and animal-related incidents.
frequency and severity, saving lives, and
accident prediction. - Predictive analytics can forecast accident likelihood and severity.
minimizing societal and economic impact.
1)Vehicle Crash Dataset Development - The dataset can guide infrastructure development and urban
2)Data processing planning by identifying high-risk areas.
3)Feature Selection - Insights can help policymakers create targeted safety regulations.
4)Use of the ID3 algorithm - Public awareness campaigns can promote safer driving practices
5)Tree construction based on common accident scenarios.
6)Model Evaluation
ACKNOWLEDGEMENT
Random forest:-When considering the use of the Figure 2.Random forest
Figure 1. Decision Tree
We are grateful to MRS.KEERTHIKA. K tree
Random Forest algorithm for traffic accident prediction,
and Dr.Narendran S M for encouraging
discussions about the topic. We acknowledge it is important to describe the specific study objectives, REFERENCE
Amrita Vishwa Vidyapeetham, Mysuru methods, and expected results. Below is a visual example
1) Indraja Smitha Tabitha Paul Chandini;(2018);A Road Accident Prediction
Campus of the proposed work using the Random Forest Model Using Data Mining Techniques;The papers suggest the use of data
algorithm for traffic accident prediction. mining technologies and algorithms for developing predictive models for road
1)Development of a complete data set accidents in India and other regions.

CONTACT 2)Feature Selection and Engineering

3)Implement Random Forest Algorithm
[2] Mrs.Kavitha Bai A.SAishwaryaThankchanE Si Krupa;(2020 may);Road
accident analysis using data miningtechniques ; The paper discuss the results of
using data mining techniques to analyse road accidents in India and also the
4)Hyper parameter Tuning accidents which might occur in the future.
Name:ANURAG S TIPPA [3] Liling Li, Sharad Shrestha, Gongzhu Hu;(2021);Analysis of road traffic fatal
Organization Name:AMRITA VISHWA 5)Evaluation accidents using data mining techniques;Association rules among variables were
VIDYAPEETHAM, MYSURU CAMPUS 6)Spatial and Temporal Analysis discovered using the Apriori algorithm
E-mail: anuragtippa757@gmail.com 7)Interpretability and Feature importance
phone no:7892817515 Figure 3. Comparison of Two
8)Compare with current model Algorithm
Figure 4.Confusion Matrix
9)Implementation and design