Machine Learning based integrated pedestrian facilities planning and staff assignment problem in transfer stations Bisheng He a,b, Hongxiang Zhang a, Keyu Wen c,d, Gongyuan Lu a,b,1 a School of Transportation and Logistics, Southwest Jiaotong University P.O. Box 610031, Chengdu, China b National United Engineering Laboratory of Integrated and Intelligent Transportation P.O. Box 610031, Chengdu, China c China Railway Economic and Planning Research Institute P.O. Box 100038, Beijing, China d School of Economics and Management, Southwest Jiaotong University P.O. Box 610031, Chengdu, China 1
[email protected], Phone: +86 (0) 138 8060 9100 Abstract Optimizing the pedestrian facilities plan in transfer stations is the problem of adjusting the facilities on the layout of pedestrian flow route and the number of machines in service to service to meet the level of services requirements. In the practice, the operation of pedestrian facilities plan is always associated with the staff assignment. Hence, we develop a machine learning based integrated pedestrian facilities planning and staff assignment optimization model in transfer stations to schedule the pedestrian facilities plan and the staff assignment together. It aims to minimize the staff assignment cost and the deviation of working time of each employee of the station. The minimizing of the deviation gains the fairness of the assignment plan. The facilities plan is enforced by the level-of-services requirement in three performance indicators including transfer capacity, transfer average time and level- of-service. The performance indicators of facilities plans are evaluated by a simulation- based machine learning method. Based on simulation results, the random forest method fits a quantitative relationship among performance indicators of the facilities plans with operation scenario attributes and facilities plan attributes.