Machine Learning Based Spatio-Temporal Prediction System for Traffic Accidents

Machine Learning Based Spatio-Temporal Prediction System for Traffic Accidents — Department of Electrical and Electronics Engineering, Bilkent University.

This project focuses on predicting traffic accidents using machine learning and deep learning techniques. We combine statistical models (e.g., SARIMAX), classical approaches (e.g., Linear Regression), boosting methods (LightGBM, XGBoost), and deep learning models (ConvLSTM) to achieve robust predictive accuracy.

The workflow includes data collection from diverse sources in Turkey (historical accident records, weather, road conditions, and points of interest), followed by preprocessing, feature engineering, training, and model evaluation. Metrics such as precision, recall, and F1 score guide optimization and fine-tuning.

The final model is deployed into DataBoss’s product, enabling proactive traffic management and improved road safety through an accessible software application.