Demand Prediction of Ride-Hailing Pick-Up Location Using Ensemble Learning Methods

Ride-hailing and carpooling platforms have become a popular way to move around in urban cities. Based on the principle of matching riders with drivers, with Uber, Lyft and Didi having the largest market share. The challenge remains being able to optimally match rider demand with driver supply, reducing congestion and emissions associated with Vehicle clustering, deadheading, ultimately leading to surge pricing where providers raise the price of the trip in order to attract drivers into such zones. This sudden spike in rates is seen by many riders as disincentive on the service provided. In this paper, data mining techniques are applied to ultimately develop an ensemble learning model based on historical data from City of Chicago Transport provider’s dataset. The objective is to develop a dynamic model capable of predicting rider drop-off location using pick-up location data then subsequently using drop-off location data to predict pick-up points for effective driver deployment under multiple scenarios of privacy and information. Results show neural network algorithms perform best in generalizing pick-up and drop-off points when given only starting point information. Ensemble learning methods, Adaboost and Random forest algorithm are able to predict both drop-off and pick-up points with a MAE of one (1) community area knowing rider pick-up point and Census Tract information only and in reverse predict potential pick-up points using the Drop-off point as the new starting point.