Prediction of Traffic Flow Based on Calendar Data on Suburban Roads (Case Study: Chalus Road)

Traffic flow prediction as a vital component in the intelligent transportation management systems can significantly improve the overall speed and flow of traffic. The purpose of this paper is to examine calendar information to maintain the active nature of traffic flow and improve the predictive capabilities of existing models for traffic flow forecasting, focusing on Chalus Road as a case study. Specifically, the Chalus road, in the mountainous regions of northern Iran, known for its sensitivity to calendar data, is examined as a case study. Three distinct models, Deep LSTM, RandomForest Regressor, and XGBRegressor, predict traffic flow using calendar-related information. These models are rigorously compared based on performance metrics. The accuracy of the prediction models is assessed, presenting actual flow values alongside predicted values and prediction errors. The results underscore the superior performance of the RandomForest Regressor model, with a traffic flow error of 86.72 vehicles. Furthermore, this model exhibits an R2 value of 0.73, surpassing other models in terms of fitting function and approximation accuracy relative to actual values. The results demonstrate that the amalgamation of machine learning models with calendar and Roads specific data to enables accurate predictions of traffic flow over time, including variations in traffic density on different days and under varying circumstances. This innovation holds substantial potential for advancing suburban road management and planning, thereby enhancing the comfort and safety of commuters.