Regression Tree Model of the Scale’s Dynamic Adjustment of Cruising Taxicab Capacity

Based on the taxi operation datasets from Taxi Information Management System in Ningbo City, a regression tree model for the scale’s dynamic adjustment of cruising taxi capacity was established by considering the average daily loading time of single taxi, mileage utilization and other indexes. The scale’s dynamic adjustment mechanism of cruising taxi capacity and thresholds standard of key indicators were proposed by coupling the functions on the balance between supply and demand and the importance ordering relationship of indicators, which consisted of the taxi ownership per person, the sharing ratio of taxi in public transport trip structure, and other indexes. The results indicate that, (1) in the three-layer structure of regression tree; mileage utilization has the strongest effect on the scale of cruising taxicab capacity. Then the average daily revenue of single taxi, the average waiting time of single taxi’s carrying, the average operation time of single taxi, and the revenue per 100 km have the stronger influence which decreased progressively. And the average daily loading time of single taxi is chosen as the third layer of classification indicator; p-values of the indicators in every layer are less than 0.05 and all passed the significance tests. The standard error and the ratio of mis-discrimination between training and testing samples are 6.13% and 0.07 which indicate the overall accuracy of model is better. (2) When mileage utilization is less than 0.6179 and the average daily revenue of single taxi is less than 798.38 Yuan, the scale of cruising taxi capacity is surplus and need to be reduced. (3) When mileage utilization is more than 0.6774 and the average waiting time of single taxi’s loading is more than 259.09 s, the scale of cruising taxi capacity is insufficient and suggested to increase 463 taxies.