Predicting Taxi Demand in Urban Areas by the Application of Hybrid Machine Learning Algorithm

抄録

This study focuses on enhancing the accuracy of the taxi demand forecasting model by location at different times of the day. The disequilibrium between taxi supply and demand occurs because both operators and taxi drivers have difficulty emotionally estimating the number of passengers and the demand from locals. Taxi driver and passenger time matching are considered the main factor for taxi allocation, but external parameters, such as the weather, may also have a strong impact. To solve this challenge, we have proposed a hybrid machine learning (HML) model. The HML, which combines linear regression (LR), random forest (RF), gradient boosting (GB), and decision tree (DT) algorithms, has been developed and tested on a year's worth of taxi and weather data in Nagaoka, Japan. Our findings show that the HML is better than any of the individual models, even if by a slight margin, which might be advantageous to taxi companies.