Deep Learning Based Research On The Demand Forecasting Method Of Shared Bikes

With the high-speed development of economy,contemporary urbans are up against many "city illnesses",such as lagging traffic management methods,imperfect urban planning,traffic jam and serious tail gas contamination.As a significant part of urban public transit system,public bikes can settle many matters such as urban traffic congestion.However,in terms of the current research on the planning of public bicycles,there is a problem that the demand of public bicycles is predicted only through experience,and there is a lack of scientific demand forecasting model to accurately forecast it.This will result in idle resources due to oversupply and users losing due to undersupply.Therefore,how to improve the forecasting method of public bicycle demand based on the existing actual data and improve the forecasting accuracy,so as to provide effective reference suggestions for the government and enterprises is an urgent problem to be solved.The demand for public bicycles is influenced to some extent by external factors such as weather,and the demand for public bicycles is a continuous time series in terms of time,which contains relevant spatio-temporal characteristics.All these factors make it possible to predict the demand for public bicycles.In this paper,a demand prediction method for public bicycles is implemented based on deep learning method.The specific work is based on the New York bicycle data set as the data source,analysis of the temporal and spatial characteristics of the quantity required for public bicycles in New York,and builds the prediction model of the demand for public bicycles according to the temporal and spatial characteristics.Considering the accuracy and efficiency of demand prediction of public bicycles,a ST-FCCNET method based on deep learning was proposed.Firstly,in this article setting up a ST-FCCNet unit construction to acquire the spatial dependence between different regions in the conurbation.Secondly,ST-FCCNET network was constructed to model the time proximity,periodicity and trend of the demand for public bicycles,so as to acquire the time dependence.Finally,combined with the influence of external factors,the final prediction results are obtained.Based on the real public bike data set,the experimental verification and analysis verify that the main factors affecting the demand of urban public bike include temperature,holidays,seasons and morning and evening peak hours.and also testify the availability of ST-FCCNet,and contrasted with the existing4 methods.The results show that the performance of ST-FCCNet is better than all other methods.