MACHINE LEARNING APPROACHES FOR TRAFFIC FORECASTING ON BLOCKED STREETS: A STUDY WITH MICROSIMULATION MODELS

Abstract

Freight deliveries on city roads with traffic signals often cause lane blockages, worsening urban traffic congestion. This issue has garnered increased attention as traffic engineers and urban planners seek sustainable ways to manage growing demand within the limits of existing road capacities. This study aims to assess a model designed to quantify the impact of freight deliveries on road capacity and delays on signalized streets in Ahmedabad. The model is based on methodologies similar to those outlined in the Highway Capacity Manual (HCM2010). The research explores how these analytical tools can be utilized in developing urban freight delivery policies. It examines delay estimation and vehicle capacity by considering various factors such as delivery locations and times, and their effects on different lanes. Machine learning techniques, including Support Vector Machines and Artificial Neural Networks, were employed to forecast vehicle capacity and estimate delays. The results demonstrate a strong alignment between the predicted outcomes and actual data.