Road Accident Analysis and Accident Severity Prediction

Abstract

Due to the exponentially increasing number of  vehicles on the road, the number of accidents occurring on a  
daily basis is also increasing at an alarming rate. With the high  number of traffic incidents and deaths these 
days, the ability to  forecast the number of traffic accidents over a given time is  important for the transportation 
department to make scientific  decisions. In this scenario, it will be good to analyze the  occurrence of accidents 
so that this can be further used to help  us in coming up with techniques to reduce them. Even though  
uncertainty is a characteristic trait of majority of the accidents,  over a period of time, there is a level of 
regularity that is  perceived on observing the accidents occurring in a particular  area. This regularity can be 
made use of in making well  informed predictions on accident occurrences in an area and  developing accident 
prediction models. In this Work, we have  studied the inter relationships between road accidents, condition  of 
a road and the role of environmental factors in the occurrence  of an accident. We have made use of data 
mining techniques in  developing an accident prediction model using Apriori  algorithm and Support Vector 
Machines. Bangalore road  accident datasets for the years 2014 to 2017 available in the  internet have been 
made use for this study. The results from this  study can be advantageously used by several stakeholders  
including and not limited to the government public work  departments, contractors and other automobile 
industries in  better designing roads and vehicles based on the estimates  obtained.