Traffic Management System Based on Density Prediction Using Maching Learning

Abstract

Traffic Management is always a daunting task, and with increasing population and number of vehicles, managing of traffic is not that easy. Now with the advancement of technologies like RFIDs, sensors, video cameras, work has been done by employing such technologies in managing the traffic wisely. But these systems only work based on traffic data captured, and there has been no intelligence in managing the traffic based on density. So, with the upcoming of machine learning which is subset of AI, there has been work for managing traffic congestion, predicting the capacity of multi-lane roads and also for determining amount of time traffic signal should be green for each junction. All these systems have been focused with developed countries where they have structured lane system and proper transportation infrastructure. In developing countries like India traffic is always on peak and voluminous due to amount of people using vehicles. So, with all these in mind, it is very much important not only to automate traffic management pertaining to congestion and traffic signal, but also to predict the upcoming traffic volume for regulating them. So accordingly, time series analysis (ARIMA) was compared with least linear regression model in terms of error and accuracy for forecasting 7-day average density traffic for 5 months. It was found that ARIMA provided the best forecasting model with reduced error and higher accuracy.