User Behavior Analysis And Demand Prediction Based On Bike-sharing Travel Data

The advancement in science and technology promotes the rapid development of urban modernization,and public transportation services are constantly innovating.Bike-sharing greatly facilitates the transportation of urban residents.The research based on big data can provide scientific basis for the management and decision-making of bike-sharing,and solve a series of chaotic phenomena in the process of bike sharing operation effectively.It is a hot topic to analyze the travel data of bike-sharing users and improve the efficiency through data mining.Bike-sharing generates a huge amount of user travel trajectory data every day.The data contains a wealth of personalized user information.It is necessary to removing null value,pre-filtering and standardizing trajectory data for repetitive,wrong,extreme and inconsistent records.The user behavior analysis adopts improved DBSCAN clustering algorithm to analyze,and obtains the rules of user travel behavior in user hot spots at different time periods.In the demand forecasting stage,the travel area is divided by using two-stage K-means clustering algorithm based on the analysis results of user behavior in hot spots.Then,the LSTM prediction model is built to predict the travel demand of bike-sharing in a certain area at different periods.The results have little error,which verifies the prediction accuracy of the model.The results of analysis and prediction provide the scientific basis for the management and decision-making of bike-sharing.It alleviates the contradiction of the mismatch between the supply and demand of bike-sharing,and reduces the of urban managers,and promotes the long-term sustainable development of the bike-sharing industry.