Abstract: For improving the frequent shortages problem of stations and the availability of the bike-sharing services,a solving method oriented to bike-sharing services scheduling was proposed to make scheduling planning in advance according to the predicted station-level bike demand.The algorithm mined the station clusters dynamically by considering both bike usage demand and geographical location between the stations in the history of riding records,and extracted the features from the multi-source data to predict the cluster-level bike demand by using XGBoost model.Then the Historical Time window K-nearest neighbor and Redistribution (HTKR) algorithm was used to estimate the station-level bike demand in different time periods to maximally meet the predicted bike demand of each station cluster.The constraint conditions of the bike-sharing services scheduling planning were established,and the ant colony algorithm was optimized by combining the chaos theory with the ant colony system to solve the process planning problem of the static scheduling of bike-sharing service.The accuracy and effectiveness of the proposed algorithm were verified by designed experiments.

Key words: sharing economy, bike-sharing services, process planning, process management, ant colony algorithm, demand prediction

摘要： 为提高共享单车服务的可用性,改善站点单车经常发生短缺的现象,提出一种面向共享单车服务调度的流程规划算法,以根据预测的站点用车需求提前进行流程规划。算法结合历史骑行记录中站点的用车需求相关性和位置关系动态挖掘站点集群,并从多源数据中提取特征用XGBoost模型预测站点集群的用车需求;通过历史时间窗口K近邻与再分配算法估计站点在不同时间段的用车需求量,以最大程度满足站点集群的预测用车需求;建立共享单车服务调度规划约束条件,结合混沌理论和蚁群系统对蚁群算法进行优化改进,解决共享单车服务静态调度的流程规划问题。设计实验验证了所提算法的准确性和有效性。

关键词: 共享经济, 共享单车服务, 流程规划, 流程管理, 蚁群算法, 需求预测