Application of Data Mining Techniques for Traffic Density Estimation and Prediction

Advanced Traveller Information Systems (ATIS) is one of the functional areas of Intelligent Transportation Systems (ITS) and it aims at providing real time traffic information to the travellers for making better travel decisions. This information would be most effective if provided to travellers during or before the start of their trip. Therefore, accurate prediction models are required in ATIS for conveying reliable information about the future state of traffic. Different methods used for the prediction of traffic parameters include historic averaging, regression analysis, Kalman filtering, time series analysis, machine learning, etc. The objective of this research is to explore the use of automated sensor data and data driven techniques for traffic state prediction under Indian traffic conditions. Travel time and traffic density (as an indicator of congestion) are used commonly to inform users about the state of a traffic system. However, these two parameters are spatial in nature and direct measurement from field is difficult. Therefore, estimation of these parameters from location based data is a challenge in many of the ITS implementations. The present study addresses the problem of estimation of traffic density with the help of location based sensors which are capable of measuring parameters such as volume and Time Mean Speed (TMS). Machine learning techniques namely, k-Nearest Neighbour (k-NN) and Artificial Neural Network (ANN) are selected as the estimation and prediction tool in this study, based on acceptable performance of the same in earlier studies.