Research On Personalized QoS-based Web Service Recommendation Method

Web service, as a new model of Web applications and distributed computing, isdeveloping very fast in recent years. However, along with the further grow of Webservice trading scale, the online trading platform structure gets more complicatedwhile it provides more and more choices. On the one hand, to the customer, with therapid growth of the type and number of Web service, the customers run into thetrouble of information overload, usually customers need to spend a lot of time to findthemselves services they need; on the other hand, to the service provider, the processthat customers browse a large number of irrelevant information and products will nodoubt make consumers submerged in the problem of information overload in thecontinuous loss. The Web service recommendation system simulates the sales staff toprovide Web services recommendation to users, it helps users to find services thatthey need, thus can complete the transaction process successfully. In the increasinglyfierce competitive environment, Web service recommendation system can save timefor users, at the same time, it also can help the trading platform to fulfill the purposesof customers’ loyalty, customers’ profitability and customers’ retention, and it canalso improve the marketable skills of the system.Quality of Service (QoS) is a key factor to conduct Web service recommendationsuccessfully. So far, there is limited research on Web service recommendation basedon QoS, and the existing research still has problems such as the low personalizedrecommendation degree and recommendation efficiency. The efficiency ofrecommendation method still needs to further improve. In view of the problems thatthe Web service recommendation system faces, the paper takes exploration andresearch of the personalized recommendation algorithm in Web servicerecommendation system.The research finds and innovation of this paper are mainlyindicated on two aspects as follows:(1) An effective personalized collaborative filtering method for Web servicerecommendation is presented. Different from the Pearson Correlation Coefficient(PCC) similarity measurement, this paper takes into account the personalized influence of services when computing similarity measurement between users andpersonalized influence of services when computing similarity between services. Basedon the similarity measurement model of Web services, this paper develops aneffective Personalized Hybrid Collaborative Filtering (PHCF) technique byintegrating personalized user-based algorithm and personalized item-based algorithm.Experimental results show that the method improves accuracy of recommendation ofWeb services significantly.(2) Combined with the characteristics that the QoS of Web service is locationsensitive, based on the personalized recommendation algorithm, this paper proposes apersonalized QoS location-aware recommendation algorithm. According to real wordWeb service dataset, this paper verifies that partial QoS attributes are highly relatedto the location of user and service. So, this paper presents a regional aggregationmodel of users and services, and in the regional model, personalized recommendationalgorithm is used. The experimental results show that, compared with the traditionalcollaborative filtering algorithm and present the most advanced algorithms, thismethod can not only make more accurate Web service recommendation for the activeuser, but also it can greatly reduce the algorithm execution time, and solve the coldstart problem of collaborative filtering algorithm, it is of great theoretical value andpractical significance.