A Comparative Analysis Of Regression And Artificial Neural Network Models For Saturation Flow Rate Prediction At Signalized Intersection Under Heterogeneous Traffic Condition

Abstract

Intersections in urban areas are critical elements of the city transport network, which has a great impact on traffic movements, traffic load and safety. The capacity to effectively predict and maximize the flow rates of saturation at these intersections is very essential in the effective control of traffic signals and the smooth operation of the movements within the cities. It is argued in this research paper that three predominant modelling methods namely, multiple linear regression (MLR), non-linear regression (MNLR) and artificial neural networks (ANN) should be compared in detail in terms of their applicability in the prediction and optimization of the saturation flow rates in the urban traffic networks. It is a field study that uses real-world field data collected in a range of urban intersections, and it carefully builds, tests and compares both statistical models and machine-learning models to check their accuracies and relate to the real world. On the basis of thorough assessment, important conclusions are arrived that not only discuss the strength and weakness of each of the modelling techniques but also provide viable ideas and suggestions to both urban traffic engineers and urban traffic policy planners. The present paper is intended to add to the cutting-edge thinking in the way of data-based solutions enhancing urban transport management and enactment of decisions.