A Comparative Study on Application of Time Series Analysis for Traffic Forecasting in India: Prospects and Limitations

Modelling of growth trend and improvement in forecasting techniques for vehicular population has always been and will continue to be of paramount importance for any major infrastructure development initiatives in the transportation engineering sector. Although many traditional as well as some advanced methods are in vogue for this process of estimation, there has been a continuous quest for improving on the accuracy of different methods. Time-series (TS) analysis technique has been in use for short-term forecasting in the fields of finance and economics, and has been investigated here for its prospective use in traffic engineering. Towards this end, results obtained from two other traditional approaches, namely trend line analysis and econometric analysis, have also been collated, underlining the better results obtained from TS analysis. A regression model has been developed for predicting fatality rate and its results have been compared with those from TS analysis. Based on the incentive provided by reduced errors obtained from using increasing number of data points for model-building, forecasting has been done for the year 2021 using time-series modelling. With most of the datasets used and locations analysed for forecasting, the TS analysis technique has been found to be a useful tool for prediction, resulting in lower estimation errors for almost all the cases considered. It has also been inferred that the proximity of the forecasting window to the sample dataset has a noticeable effect on the accuracy of time-series forecasting, in addition to the amount of data used for analysis.