Machine Learning Approaches to Traffic Accident Analysis and Hotspot Prediction

Traffic accidents are one of the most important concerns world, since they result in numerous casualties, injuries, and fatalities each year, as well significant economic losses. There many factors that responsible for causing road accidents. If these can be better understood predicted, it might possible to take measures mitigate damages its severity. The purpose this work is identify using accident data from 2016 2019 district Setúbal, Portugal. This aims at developing models select a set influential may used classify severity an accident, supporting analysis on data. In addition, study also proposes predictive model future based past Various machine learning approaches create models. Supervised methods such decision trees (DT), random forests (RF), logistic regression (LR), naive Bayes (NB) used, unsupervised techniques including DBSCAN hierarchical clustering. Results show rule-based C5.0 algorithm capable accurately detecting relevant describing Further, results suggests RF could useful tool forecasting hotspots.