A Model Stacking Approach for Ride-Hailing Demand Forecasting : a Case Study of Algiers

Abstract

A good understanding of supply of taxis drivers and demand of passengers through the forecasting of future demands is important for an intelligent transportation system in smart cities. A good prediction allows a better allocation of taxi fleets, reduce passengers' waiting time and energy waste for taxis. In this paper, we harness the power of statistical and machine learning models in a joint ensemble model and propose a stacking approach which combines the predictions of ARIMA, SARIMA, MLP, LSTM and XGBoost. Our proposed approach consider also external factors such weather conditions and national holidays. We consider in this work the city of Algiers as a case study, using ride hailing data of an Algerian ride-hailing platform. Experimental results show that our model performed better than the selected baseline statistical and machine learning algorithms.