Among various public safety challenges, traffic accidents are one of the concerns. According to NHTSA, traffic fatalities for 2019 shows that 36,120 people have died in motor vehicle traffic crashes. Therefore, predicting the accident risks is really important. And if the traffic car accident risks can be predicted for specific conditions under which they are happening then it will help our local government and law enforcement agencies in taking prevention measures so as to mitigate or avoid those accidents in future. In the present research various impact of different categories of data attributes was analyzed and top contributing features were found for modeling. Various machine learning models have been experimented like Logistic Regression, Multi Layer Perceptron, Random Forest Classifier, XGBoost in order to find out the best accuracy prediction of accident risk.