Road accidents are caused by several factors. Taking this and the shortcomings of prior research studies on accident prediction in India into account, we propose an accident prediction system that will utilize the data collected by Mumbai Traffic Police, along with static data of Mumbai roads and weather data obtained from various sources. Several Machine Learning Models have been applied to our prepared dataset and a mean accuracy of approximately 85% is obtained. The XGBoost algorithm gives the maximum accuracy of 86% among all the models. It is observed that the features: Latitude, Longitude, Date and Time, Shape Length of a Road, Pressure, Wind speed, and Humidity affect the severity of accidents the most. The best model obtained is integrated into a real-time mobile application that provides drivers and road commuters with a percentage accident chance for each road on the user's route. The proposed system thereby will alert drivers and commuters, serving as a safety measure and helping Road Authorities and Governments gain more insights into accidents and prevent road accidents in the future.
Safe Routes Recommendation for Drivers by Real-Time Prediction of Accident Risk Scores
23.09.2022
1073525 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Predicting Accident Risk in Novice Drivers
British Library Conference Proceedings | 1995
|Congestion on Risky Routes with Risk-Averse Drivers
British Library Conference Proceedings | 2005
|Accident Prediction Models for Ontario Truck Drivers
British Library Conference Proceedings | 1996
|Signal intersection traffic accident risk real-time prediction method and system
Europäisches Patentamt | 2021
|