Forecasting car rental demand based temporal and spatial travel patterns

Recent years, shared mobility services have gained momentum across the world. Meanwhile, rental car industry has seen great developments in China and has reached a scale of economy. Knowing the rental behavior pattern and forecasting the demand become more important for rental businesses. To this end, in this paper, we aim to analyze the rental mobility pattern by examining multiple factors in a holistic manner. A special goal is to predict the demand of a given region. Specifically, we first analyze regular mobility based on real trips of rental cars. Then, we extract key features from multiple types of rental-related data, such as rental behavior profiles and geo-social information of regions. Next, based on these features, we develop a multi-task learning based regression approach for predicting rental cars' demand. This approach can effectively learn not only fundamental features but also relationships between regions by considering multiple factors. Finally, we conduct extensive experiments on real-world rental trip data collected in Beijing. The experimental results validate the effectiveness of the proposed approach for forecasting rental demand in the real world.