Research On Web Service Performance Prediction Methods Based On QoS

After the World Wide Web appears, computers across the world are linked together, the data resources of these computers make up a large-scale database. Faced with such a large database, looking for the information you really want is no different from fishing pin in the sea. How to find the real useful information hidden in large amounts of data, only relying on traditional statistical methods and information retrieval methods can not solve above problems. Right in the above background, Data Mining technology comes into being. Getting the most valuable information resources from the large amounts of data quickly and efficiently is important in today’s time. Thus, the techniques and tools related to the data mining are facing good prospects. Among them, the classification prediction technology as an important means of intelligent decision-making will play an important role in intelligent systems, and how to predict and judge its performance for a new Web service quickly and accurately is becoming a hot and difficult issue in the current research field.In the country863projects-the Environmental perception and evolution technology of Internetware based on layered adaptation strategies, to address the optimization and upgrade of composition service environment self-adaption operating and supporting platform, and based on the QoS Got in the Mining stage, this thesis proposes Web service performance prediction methods based on QoS. According to the development direction of current classification algorithm, and against two kinds of different knowledge modes-association rules and emerging pattern, this thesis study and design two kinds of service performance prediction methods. In the methods, give several QoS measure standards and definitions respectively, study new rule cutting algorithms, and propose effective and rational scoring standard classification strategies, provide valuable ideas and improvement for predicting performance and category of new Web service instance accurately and effectively.This thesis firstly studies the Web service performance prediction framework based on QoS, and designs the basic process of performance prediction methods according to the main parts of prediction methods. Secondly, this thesis proposes the Web service performance prediction method based on FeedbackAccuracy(CBFA), gives the definitions of the FeedbackAccuracy and IntegratedValue, proposes thoughts of the "father and son" relationship, priority and optimal rules, studies new rule cutting algorithms, and gives new solution strategies in the aspects of scoring standard and classification, then proves the CBFA has better classification accuracy by experimental analysis. Finally, this thesis proposes the Web service performance prediction method based on Emerging Pattern(CBEP), gives the prediction model and prediction process of the algorithm CBEP, improves the method of determine problem of the minimum growth rate threshold, and selects different minimum growth rate thresholds for each category. In addition, in the aspect of scoring standard, this thesis puts forward the idea of standardization, makes the performance prediction for unknown service more reasonable and accurate.