Predicting the Next State of Traffic by Data Mining Classification Techniques

Abstract

Traffic prediction systems can play an essential role in intelligent transportation systems (ITS). Prediction and patterns comprehensibility of traffic characteristic parameters such as average speed, flow, and travel time could be beneficiary both in advanced traveler information systems (ATIS) and in ITS traffic control systems. However, due to their complex nonlinear patterns, these systems are burdensome. In this paper, we have applied some supervised data mining techniques (i.e. Classification Tree, Random Forest, Naïve Bayesian and CN2) to predict the next state of Traffic by a categorical traffic variable (level of service (LOS)) in different short-time intervals and also produce simple and easy handling if-then rules to reveal road facility characteristic. The analytical results show prediction accuracy of 80% on average by using methods