Travel Distribution Prediction Model for Bike Sharing Considering Congestion Index

Accurate prediction of bike-sharing trip distributions is critical for urban non-motorized traffic planning and bike-sharing operation scheduling. This paper takes residents' travel destination decision-making behavior as a starting point and proposes a single-factor prediction model for bike-sharing trip distribution considering POI. Based on this model, a two-factor prediction model considering congestion index and its improvement model are developed. Based on bicycle riding data from Futian District, Shenzhen, this study analyzes the clustering phenomenon within the travel OD network. It utilizes a community detection algorithm from complex networks to partition Futian District into four traffic analysis zones (TAZ). The influence of POI, congestion index and travel distance on bike sharing are then analyzed, and it is found that the number of POIs shows a linear positive correlation with the amount of bike sharing trips. At the same time, the congestion index shows a significant positive correlation with the amount of bike sharing trips, especially in the areas with larger travel volumes, whenever the congestion index increases by 0.1, the proportion of bike sharing trips will increase by 6%~7%. The travel distance shows a long-tailed logarithmic distribution characteristics. The prediction results show that during weekends, the accuracy of the two-factor improvement model developed in this paper is 81.2%, 79.5%, 80.1%, and 78.9% for the four TAZs, respectively. During weekdays, the accuracy rates were 78.7%, 76.3%, 80.8%, and 75.5%, respectively. Compared to the radiation model, the prediction accuracy was improved by a maximum of 51.1%. © 2024 Science Press. All rights reserved.