Hybrid Simulated Annealing and Random Forest for Traffic Density Prediction in VANETs

Abstract

The study addresses the issue of predicting traffic density in Vehicular Ad-hoc Networks (VANETs), where dynamic and unexpected traffic patterns limit accurate forecasting. Recent models frequently encounter challenges with accuracy caused by overfitting or complications in handling real-time data. The study introduces a hybrid model that combines Random Forest with Simulated Annealing, optimising the model’s parameters to mitigate overfitting and improve reliability. The research follows several steps: first, data from a VANETs dataset was collected and preprocessed, and then several standard machine learning models, like Linear Regression, Decision Trees, Random Forest, Support Vector Regression, and K-Nearest Neighbors, were tested. The Random Forest model showed the best performance metrics and was optimized using Simulated Annealing. The hybrid Simulated Annealing-Random Forest model significantly improved accuracy, outperforming traditional models.