Supply-demand Forecasting For a Ride-Hailing System
Ride-hailing or Transportation Network Companies (TNCs) such as Uber, Lyft and Didi Chuxing are gaining increasing market share and importance in many transportation markets. To estimate the efficiency of these systems and to help them meet the needs of riders, big data technologies and algorithms should be used to process the massive amounts of data available to improve service reliability. The model developed predicts the gap between rider demands and driver supply in a given time period and specific geographic area using data from Didi Chuxing, the dominant ride-hailing company in China. The data provided includes car sharing orders, point of interest (POI), traffic, and weather information. A passenger calls a ride (makes a request) by entering the place of origin and destination and clicking “Request Pickup” on the Didi phone based application. A driver answers the request by taking the order. Our training data set contains three consecutive weeks of data in 2016, for large Chinese city which is referred to as City M. Though the training set is relatively small when compared to the whole of Didi's ride sharing market, it is large enough so that patterns can be discovered and generalized. These data were made available to researchers and entrepreneurs by Didi after removal of some identifying information.