Traffic Demand Prediction of Urban Public Bicycles with the Consideration of Land Use

Aiming at the problem that newly built public bicycle stations cannot predict future use demand based on historical data, a modified geographically weighted regression model was proposed to explore the relationship between demand generation and accessibility at unit time nodes. The modified model took road network distance as the constraints to access the overlapping attraction area of docking stations based on Thiessen Polygon. Meantime, in order to reduce the prediction error caused by land location, the modified model added land mix degree and building strength as explanatory variables. The proposed model was utilized to analyze data from the docked bike-sharing system in Xi'an. The results indicate that the production rates of various land types are found to be maximum in the morning and evening peak hours with variable patterns in daily change. The production rate of bike-sharing system will gradually decrease as the distance between origin/destination and target docking stations increases, showing linear attenuation in the morning peak period, exponential attenuation in the evening peak period and cubic attenuation in the non-peak period. The findings can be utilized to determine the location and scale of new docking stations of bike-sharing system in Xi'an and predict relative usage demand rates. © 2022, Editorial Department, Journal of South China University of Technology. All right reserved.