Machine Learning for Predicting Traffic and Determining Road Capacit

This study proposes the use of machine learning techniques to predict traffic speed based on traffic flow and other road-related features, utilizing the California Freeway PeMS traffic dataset. Extensive research has been dedicated to the prediction of road speed; however, the primary challenge lies in accurately forecasting speed as a function of traffic flow. The learning methods compared include linear regression, K-nearest neighbors (KNN), decision trees, neural networks, and ensemble methods. The primary objective of this research is to develop a model capable of estimating road capacity, a crucial factor in designing an auction system for road usage. The findings reveal that the performance of each algorithm varies with the selection of features and the volume of data available. The results demonstrate that ensemble methods and KNN surpass other models in accuracy and consistency for predicting traffic speed. These models are then employed to create a flow-speed graph, which aids in determining road capacity.