A Study on Predicting Traffic Flow by Random Forest Based on Treatment of Eliminating Skewed Distribution

Abstract

With the accelerating urbanization and complexity of transportation networks, short-term traffic flow prediction has become increasingly crucial in traffic management. Addressing this issue, this study introduces a random forest model for traffic flow prediction, incorporating deskewing treatment. Quantile-Quantile Plot analysis reveals the right-skewed distribution of traffic flow data, necessitating the use of a logarithmic function for deskewing. Further analysis indicates temporal patterns in the data. To validate the deskewing effect and model performance, Lasso, SVR, and Random Forest regression models were employed. The results demonstrate that deskewing significantly enhances prediction accuracy, with the Random Forest model outperforming others at various junctions.