Research On Service Recommendation For Crowdsourcing Distribution

In the emerging field of shared economy,driven by the last three kilometers of the delivery service area also began to try crowdsourcing logistics,the model to APP as a platform,based on the broad social resources to provide users with convenient and efficient distribution services,and customer satisfaction is one of the important means of competition for crowd logistics platform to maintain the user and open up the market,which also makes it from a large number of social resources to obtain the right information to meet the needs of personalized distribution services to become the crowdsourcing distribution needs to address the key issues one.In this paper,combining the National Natural Science Foundation of China(71471165): “Research on Personalized Information Service of Context-Awareness Driven by Socialization in Ubiquitous Computing Environment.”,this paper proposes a collaborative filtering recommendation method based on multi-attribute multi-attribute and multi-temporal scoring of distribution service.At the same time,it explores the personalized recommendation in Crowdsourcing distribution for customer satisfaction Impact.The main contributions of this paper include:(1)Propose the personalized delivery service for the target userBased on the existing crowdsourcing logistics service model,this paper proposes a distribution service based on user preference for the existing distribution service,which mainly studies which distribution scheme is suitable for the target user,and combines personalization and logistics service to solve the problem of distribution service quality in the development of package Logistics.(2)Analysis on the scoring behavior of users’ distribution serviceThe user’s score fluctuation analysis is performed based on a service attribute,and the quantified value is used to score prediction and multi-attribute weight determination of the service attribute for the target user,laying a foundation and foreshadowing for the following research work.(3)Target user makes the nearest neighbor scoring predition based on a propertyThrough the construction of the "user-distribution service" scoring matrix,all users score the rating of all the distributed service,and obtain similar users by calculating the similarity value,and then forecast the target user’s property rating by the nearest neighbor algorithm.(4)Constructs the multi-attribute multi-time-series forecast model of the distribution service scoreFirstly,the paper transforms the traditional logistics scoring model from the total score to the multi-attribute scoring of the distribution service.Through the analysis of the target users’ historical scoring behavior,this paper searches for the distribution service attributes,which is the target user’s preference.Based on the analysis of three and four chapters to build the target user to the distribution service attributes score prediction method Then through the introduction and comparison analysis of the multi-attribute weighting method,the total score of the target service to the distribution service is obtained.(5)Experimental analysis and case studyFirstly,the method of weight determination is analyzed and compared,and an appropriate weight determination method is obtained.Based on that,the following experiments are carried out.Based on the construction of service attribute system and the method of score forecasting,this paper evaluates the accuracy of the algorithm(presicion),the absolute mean variance(MAE)and the recall rate by comparing the user satisfaction before and after the algorithm.Then we use the case of Eleme as background to find out that the collaborative filtering recommendation algorithm based on the multi-attribute scoring of crowdsourcing logistics service can effectively improve the service satisfaction of crowdsourcing platform for enterprises,and reduce the cost of developing customers and maintaining customers Significance.