Using Historical Trip Information to Determine the Waiting Time Required for Taxi Services

Taxi is a standard mode of transportation in all countries. Whenever a customer requests a taxi service, a waiting time is estimated to indicate how long the client will wait to be picked up. This paper develops prediction models that predict the waiting time for taxi services. These models have been used to develop historical trip information that does not include vehicle GPS coordinates, traffic information, and weather data. Verification has been performed by calculating the accuracy percentage for the classification Random Forest and k-Nearest Neighbours algorithms and the Root Mean Squared Error (RMSE) for the regression Random Forest model. Validation has been done on a new dataset by comparing the predictions with the actual waiting time. The same validation techniques have been used with the waiting time estimated by eCabs. However, the validation results show that the estimated waiting time by eCabs is slightly more accurate than predicted. Integrating GPS vehicle locations, traffic and weather data from a business application would be helpful and should significantly improve the prediction accuracy.