Research And Application Of Web Service Recommendation System Based On Semi-supervised Clustering And Meta-path Based Link Prediction

With the rapid development of the Internet and the recommendation technology,the recommendation system gradually presents many new features,which brings new challenges for the recommender system.Web service recommendation has many intelligent features.It aims to meet the requirement of potential users and takes the active recommendation methods making the people transfer from the active searcher to the receiver.It effectively alleviates the problem of information overload,and thus attracted more and more attention from the researcher.Compared with the traditional service recommendation,current recommender systems become more socialized and complex.It needs to accurate extract the situational factors and preference of the users,then to improve recommendation results effectively.However,the current service recommendation methods only consider the influence of the context information and ignore the users' social relationship.This relationship can improve the performance of the recommender system.Therefore,it has become an important topic in the field of recommendation to combine the context information and social factors.Based on the National Natural Science Foundation "Social Service Recommendation Based on Heterogeneous Spatial Information Networks in Mobile Environment”,this paper introduces the idea of heterogeneous spatial information networks,to mine the potential relationship between the parties in depth through semi-supervised clustering and meta-path link prediction algorithms,and integrate some other recommendation strategies,and then improve the performance of web service recommendation.The main works and contributions of this paper are as follows:(1)This paper analyzes the background of the existing Web services,analyzes the present situation of the research on the social network and the heterogeneous spatial information network,and also describes the organizational structure of the paper.(2)Based on the analysis of the existing Web service recommendation technologies,we analyze the characteristics of each participant in the heterogeneous network,and then measure the relationship between the involved nodes,construct a heterogeneous spatial information network based on the users and services.(3)This paper proposed a novel recommendation approach based on the semi-supervised user clustering and meta-path based link prediction,conduct the measurement of the relationship between users,and calculate the relationship between the users and services,services and services,users and users in heterogeneous spatial information network.After that,we compute the link probability between every element of different meta-paths through random walk algorithm.Compared with the traditional recommendation algorithms,the effectiveness of the proposed algorithm is verified by the experiment.(4)We analyze the problem of the Web service recommendation,describe the requirements of the Web service recommendation system,and also design the Web service recommendation system based on semi-supervised clustering and meta-path based link prediction algorithm.In addition,the paper gives a detailed description for the requirements and processes of functional modules.(5)We perform the prototype of Web service recommendation system by using the Java language and MySQL.In order to give an intuitive description for the system,we give some important parts of our Web service recommendation system,and then test each part functional modules.