Collaborative Filtering Qos Prediction Based On K-means Clustering

Web service QoS prediction has become a popular research field in service computing.Collaborative filtering is one of the most commonly used techniques in QoS prediction,and it is widely used in personalized QoS prediction algorithms.The traditional collaborative filtering based QoS prediction algorithm has two major key issues has not yet been effectively addressed.First of all,QoS prediction algorithm is based on the user's historical observation data,the prediction of the effective need to historical data to the credibility of the premise.Second,due to the dynamic characteristics of QoS,time-aware prediction methods are more needed.In order to solve the above two key problems,this thesis proposes two kinds of collaborative filtering QoS prediction methods based on K-means clustering algorithm for static and dynamic datasets,static trustworthy QoS prediction method based on two-phase K-means clustering and dynamic time-aware QoS prediction method based on K-means clustering.Among them,the first method solves the problem that there is untrusted user on the static data set,which leads to the decrease of prediction precision.The second method solves the problem of dynamic time perception prediction under sparse matrix.Finally,two QoS prediction methods are evaluated separately based on the static and dynamic data sets,respectively.Compared with other classical QoS prediction methods,the prediction accuracy of the method 1 on two static data sets is improved by 13%and 23%,respectively,and for the method two,the prediction accuracy is improved by 7%and 37%on the two dynamic datasets,respectively.