Analysis And Prediction Modeling Of User' Features Based On Carsharing Big Data

The carsharing is the main pattern of the sharing economy for the traffic travel.It is the key method to realize the green travel.Through the carsharing travel pattern,the vehicle using efficiency would be improved and the car ownership can be decreased,which have contribution to saving travel resources and environmental protection.The carsharing travel pattern attracts the increasingly focus and application.The development of carsharing significantly depends on the development of carsharing companies.Establishing reasonable operating strategy could improve the benefit of carsharing companies,which further ensure the development of carsharing companies and the carsharing travel pattern.However,in the current process of carsharing development,the carsharing companies need to face the problem of benefit anxiety,which is the contradiction between high cost and low benefit.The problem of benefit anxiety derives from the lack of understand to the car usage and travel behavior of carsharing users.Therefore,analyzing the car usage and travel behavior of carsharing users and establish corresponding models has great significance.The existing literatures mainly focus on the aspect of carsharing development and analyze the demand and purpose for carsharing users.However,few studies consider the development of carsharing companies to explore the relation between company benefit and behavior of carsharing users.The company benefit is mainly from the car rental behavior of users.Therefore,to improve the benefit of carsharing companies,this study analyzes the car usage and travel behavior of the carsharing users with different benefit contribution,and attempts to discover the difference of behavior between carsharing users with different benefit contribution.The result can provide the valuable suggestions with companies to determine the user-based operating strategy.In order to explore the relationship between company benefit and user behavior,based on the order data from carsharing operation,the characteristics of carsharing users are analyzed.The dataset includes more than 100,000 data points,which records the travel behavior of 5,202 users.Firstly,by using the two-step clustering method,the new users are classified into three groups,which include high benefit contribution users,medium benefit contribution users and low benefit contribution users;Secondly,analyzing the characteristics of carsharing users during 84 days and exploring behavior characteristics by using the description and multiple logistic method.The results reflect the distribution pattern of the car rental time,position,vehicle modes,car rental duration and distance of carsharing users,and the difference between bahevior of users with different benefit contribution;Thirdly,based on the multilayer perceptron neural network estimation model,the value of the users during three months is analyzed.The model uses the multiple dimensions statistical data as the input,and further predicts the user values during 84 days through minimum observation period.The results indicate that the user values during 84 days can be determined by using the data of five weeks.The predicting accuracy can reach larger than 85%.This study analyzes the car usage behavior and establishes the predicting models,which has positive effects to improve the benefit of carsharing companies and development of carsharing travel pattern.