Traffic Forecasting

Abstract

The main objective of this study was to use machine learning and XAI (Explainable artificial intelligence) to predict and explain the traffic state of a road network. As a result, a case study was presented that uses the XGBoost algorithm to predict the amount of traffic on a specific road of Athens during the next two hours. Furthermore, the SHAP framework was used to explain the model and help the user understand the forecasts that were made. Moreover, in order to improve the accuracy of our model weather data was added from Copernicus, the European Union’s Earth observation program [23]. 26 variables were used to create the model, which were divided into three groups: time-based factors, weather variables and function variables.