Usage Demand Forecast and Quantity Recommendation for Urban Shared Bicycles

Abstract

The rise and wide use of urban shared bicycles brings great convenience to residents' daily trips. However, as the number of shared bicycles continues to increase in the city, more and more shared bicycles are being allocated unreasonably, which has a great impact on the public transportation and city appearance. To solve this problem, in this paper, we focus on the recommendation of the number of shared bicycles near the subway station and take the volume of the station's outbound passenger flow as the potential demand to study. Along this line, we first collect and investigate the subway traffic data which are gathered by automated fare collection(AFC) systems and some related feature data as the basic data source for the demand prediction. Second, upon these data, we develop a novel passenger flow forecast model based on an advanced Xgboost method and the idea of sliding window. Combined with the predicted results, we then design a method to transform the passenger flow results into an usage demand index and recommend a suitable number of shared bicycles for a specific subway station by it. Finally, we conduct a series of experiments with a variety of real-world data, and the results demonstrate our model can achieve a better performance and give us a reasonable quantity recommendation.