Accident Severity Unveiled: How Weather, Roads, and Time Conspire in Machine Learning Predictions

Traffic accidents pose a significant socio-economic problem; thus, it is vital to predict factors that might influence severity to prevent potential risks. Based on the above aspects, this paper aims to design a model for the machine learning assessment of traffic accidents and severity based on weather, time, and location. These include the numerical such as the temperature and wind speed while the other is categorical such as weather conditions. For missing values and categorical inputs, the pre-processing step is taken, for temporal and interaction features the feature engineering step is performed and out of all these steps the most crucial factors for the accident severity are filtered from all the variables. We benchmark five machine learning models with respect to three specific attributes, among which there is accuracy. This study proves that XGBoost has the fastest and highest accuracy in classifying the level of seriousness of the accident. Such a discovery opens up the possibilities of applying machine learning methods at the higher advanced level to improve traffic safety. It is crucial to point out that this research pays great attention to the application of extended analytical models and big data in increasing traffic safety and outlining prevention measures.