Real-Time Traffic Prediction Using Neural Networks

Abstract

Somewhat recently, the issue of traffic has become more extreme because of industrialization particularly in significant urban communities. A higher standard of living in cities is pushing people to utilise their own vehicles instead of public transportation. Thus, there is tremendous increase in traffic, which leads to a variety of other associated issues. In order to get from one place to another, one must spend more time travelling than they would have otherwise, which also raises the fuel consumption. The traffic data that is gathered by various detectors is highly dynamic and non-stationary. For example, the number of vehicles turning at an intersection cannot be stated with greater accuracy. Consequently, it is challenging to develop a mathematical model to determine green time. However, the introduction of Intelligent Transportation Systems (ITS) in modern times enables the detection of traffic events, communication, information processing, and user action. One of the most crucial criteria for this system to work successfully is the ability to precisely predict the pattern of the traffic stream. In order to predict future traffic flow, a system that uses artificial neural networks and real-time traffic data was proposed in this study. Neural networks have the ability to predict the future by learning from the past. Specifically, this study predict traffic volume in two levels such as short-term for every 15 minutes interval and mid-term for every one hour of a given day. These predictions help in dynamic traffic management applications like signal control, congestion management, and travel time predictions etc.