Research On Multi Factor Demand Forecasting And Scheduling Method Of Urban Public Bicycle

Since the 1960 s,the emergence of public bicycles has become the first choice for urban residents to solve short-distance travel and green travel.At the same time,it has also become the best choice for alleviating traffic pressure and innovating public transportation.However,due to the accelerating pace of life,problems such as traffic congestion,unbalanced supply and demand of bicycle stations,and delays in vehicle dispatching have become more prominent.There has been a phenomenon that there are no cars to borrow and no place to pay for individual stations in the bicycle system.It has greatly affected the lasting development of the public bicycle system and the user’s travel experience,and it has also become a major problem faced by traffic managers and operators.This article analyzes the status information of bicycle stations and the user’s cycling data,based on various travel characteristics,predicts the bicycle demand at different stations,and proposes new ideas for bicycle scheduling schemes,which provide for the rebalancing of the urban public bicycle system.An important theoretical basis.First of all,the paper expounds the development status of urban public bicycle systems at home and abroad,site planning,bicycle demand forecasting,and scheduling between sites.It focuses on the problems of imbalance between supply and demand in the current public bicycle system,and the singleness of scheduling delay.Describes the research methods and related theories to solve this problem,including K-means clustering,the algorithm ideas of row sampling and column sampling in random forest,XGBoost algorithm and ant colony algorithm,etc.Secondly,based on the trip data,station distribution,meteorological conditions and other information of the public bicycle system in the San Francisco Bay Area,the analysis focuses on the influence of time factors,geographic location,meteorological conditions,and the relevance between stations on the amount of urban public bicycle trips.influences.According to the regionality of bicycle stations and the time-varying demand of bicycles,a demand forecasting model for public bicycle stations based on XGBoost and K-means clustering is proposed to accurately and quickly predict the demand for urban public bicycles.Taking into account the importance of time and location factors,use it as an important feature and apply different weights to feature-weighted K-means for site clustering.Combining the use of random forest row sampling and column sampling methods to prevent overfitting,improve The XGBoost algorithm acts as a predictor to improve the prediction accuracy of the prediction model.Experimental results show that the demand forecasting results of this method are better than other methods,and the error rate is 8.2% higher than that of random forest,which lays a solid foundation for the future scheduling scheme.Finally,the bicycle scheduling problem is described,several existing scheduling methods are explained,and the current bicycle scheduling problems between the public bicycle system stations are analyzed.Aiming at the delay and singularity of scheduling,based on the accurate prediction of the prediction model,the status of public bicycle stations can be accurately grasped in real time.According to the actual distance between public bicycle stations in the San Francisco Bay Area,the scheduling problem is modeled and analyzed.Finally,The improved ant colony algorithm based on price incentives is used to simulate the scheduling of this problem,and the scheduling scheme based on the user’s independent choice is realized.The experimental results show that when the initial unit price reaches 0.35,the load balance of the site reaches 86%,which greatly eases the rental pressure of the site and basically realizes the rebalancing of the bicycle system.