Short Term Prediction of Traffic Parameters Using Support Vector Machines Technique

Accurate and precise prediction of traffic variables such as speed, volume, density, travel time, headways etc. is important in traffic planning, design, operations, etc. Short term prediction of these variables plays a very important role in Intelligent Transportation Systems (ITS) applications. Under Indian scenario, this short term prediction of traffic variables has gained greater attention with the recent interest in ITS applications such as Advanced Traveller Information systems (ATIS) and Advanced Traffic Management systems (ATMS). In the context of prediction methodologies, different techniques such as time series analysis, statistical methods, filtering techniques and machine learning techniques have been suggested in different studies in addition to the historic and real time approaches. However, for traffic conditions such as the one existing in India, with its heterogeneous and less lane disciplined traffic, many of these techniques may not bring the accuracy that was reported in literature under homogeneous traffic. There are only very limited studies on the application of these techniques for traffic conditions such as the one existing in India. The present study proposes the application of a recently developed pattern classification and regression technique called support vector machines (SVM) for the short-term prediction of traffic variables under mixed and less lane disciplined traffic conditions. An ANN model is also developed and a comparison of the performance of both these techniques is carried out.