Predictive Modeling: An Attempt at Predicting Travel Times In Bengaluru Accounting For Geographic And Economic Effects

Abstract

Abstract The transport system of a country reflects the efficiency and growth of the country. As population increases, the number of vehicles increase, congestion and traffic increase leading to increase travel times, Evolution comes about in the transport system of a country, to increase physical connectivity and economic development, to reduce congestion and travel times. This paper aims to use machine learning algorithm on big data to understand how the effects of rainfall, temperature and pollution help predict travel times. The four machine learning algorithms used include linear regression, ridge regression, random forest regression and elastic net regression. The predicted travel times obtained by all models were compared with the observed travel times in order to determine which model gives better prediction. From the predictive modeling algorithms run on these datasets it is observed that, random forest regression is best suited in predicting travel times in Bengaluru City from ith zone to jth zone in the pth hour of weekdays and weekends after accounting for effects of pollution, temperature, rainfall and economic activity.