BIG DATA FOR PREDICTIVE ANALYTICS OF TRAFFIC VIOLATIONS

This paper produces observations of the impact of the traffic offence on traffic accidents by using big data analysis techniques dependent on traffic violations information that was updated daily in Montgomery County within the USA. By knowing the reason for traffic accidents, the aim was to recognize infringement. Also, it conceives to use the data to compute protection premiums for insurance agencies. The aimed hypothesis is to predict future violations leading to an accident, predict accident fatality, i.e., personal injury or property damage. Additionally, it attains to utilize the data to calculate insurance premiums for insurance corporations. Several classifications, clustering and regression models are considered in our analysis, like Single Tree, Random Trees, K- Means clustering, multiple regression, and Naive Bayes. The first and second model considers Random Trees and Single Trees as the best algorithms per our business case due to the importance of high sensitivity and a high F1-score. Model III considers K-means clustering and Single Tree classification the best algorithms for having the ability to produce clusters with their numerous violation types and count of injuries.