高性能计算导论课件1数据挖掘.pptVIP

Distance matrix between n landmarks and w Web service hosts min Squared sum of error Regularization term Web service host Approach 4: Network Coordinate Based Web Service Coordinate Computation min Service user Web service hosts Historical data Reference information of landmarks Available historical data constraints Regularization term Approach 4: Network Coordinate Based Service User Coordinate Computation The set of Web services with unknown response time data The coordinate of service user u The coordinate of Web service si Approach 4: Network Coordinate Based QoS Prediction Result Performance Comparison Approach 5: Ranking-Based Prediction Select the optimal Web service from the candidates Neighborhood-based approaches:　Predict QoS values ? rank the candidates Ranking-based approach: Rank the candidates directly without predicting QoS values Expected values: (2, 3, 5) on (ws1, ws2, ws3) Prediction 1: (3, 2, 4); MAE = (|2-3| + |3-2| + |5-4|)/3 = 1 Prediction 2: (1, 2, 3); MAE = (|2-1| + |3-2| + |5-3|)/3 = 1.3 Ranking-based: Expected Ranking: ws1 ws2 ws3 Prediction 1: ws2 ws1 ws3 Prediction 2: ws1 ws2 ws3 Approach 5: Ranking-Based Prediction User preference on two Web services which have been invoked previously: User preference on pairs of Web services that have not both been invoked by the current user: Approach 5: Ranking-Based Prediction Kendall Rank Correlation Coefficient (KRCC) N is the number of Web services. C is the number of concordant pairs between two rankings. D is the number of discordant pairs. Target Web services: (ws1, ws2, ws3) User 1 observed response-time: (2, 3, 5) User 2 observed response-time: (1, 2, 3) User 1: ws1 ws2, ws1 ws3, ws2 ws3 User 2: ws1 ws2, ws1 ws3, ws2 ws3 N = 3; C = 3; D = 0; Sim(user1, user2) = (3-0) / (3(3-1)/2) = 1 Target Web services: (ws1, ws2, ws3