Research On Some Key Technologies In Service Calculation

With the rapid development of modern service industry and the service economy, service chain has become a pillar industry of national economy, as an important part of the service industry chain, the choice of service and found that face enormous challenges, the dynamic, open and variety of the uncertainty in network environment, such as false, malicious feedback, service QoS dynamic fluctuations, etc.), makes the service selection result and actual result, the deflection service selection is difficult to get a long-term and stable quality, eventually leading to failure service selection., on the other hand, in the face of network services index increased in quantity, the existent service selection methods are too much to consider in the process of planning services and ignore the global optimal computation time or time complexity to the attention of, and thus often leads to a composite service in consumption during the operation of huge calculation time, this is unacceptable in practice.Therefore, how to efficiently correct selection service is a key problem affecting the development of services science in depth. Service recommendation technology is should choose and discovery challenge.The innovation of the thesis points as follows:A, Service recommendation algorithm based on user clustering (UCC-CF)This paper put forward a new cloud service QoS valuation framework system, designed to provide personalized user-centered service recommendation system, and provide personalized service QoS value is the basis of service recommendation.Considering the inherent dynamic cloud services, distribution, the characteristics of virtualization, the paper assumes that:with the same call history information of users (for example, distribution range, the development environment, network quality) has the same quality of service.At the same time, for cloud service providers, the same user behavior has the same service QoS.The core part of the paper is based on the clustering algorithm to the user and service divided into groups according to different users, through these groups to forecast the unknown service QoS.In UCC-CF algorithm is put forward by the user after completion of clustering, using collaborative filtering algorithm to predict the QoS of clustering values, which are recommended.This paper put forward a new cloud service QoS valuation framework system, designed to provide personalized user-centered service recommendation system, and provide personalized service QoS value is the basis of service recommendation.Considering the inherent dynamic cloud services, distribution, the characteristics of virtualization, the paper assumes that: with the same call history information of users (for example, distribution range, the development environment, network quality) has the same quality of service.At the same time, for cloud service providers, the same user behavior has the same service QoS.The core part of the paper is based on the clustering algorithm to the user and service divided into groups according to different users, through these groups to forecast the unknown service QoS.In UCC-CF algorithm is put forward by the user after completion of clustering, using collaborative filtering algorithm to predict the QoS of clustering values, which are recommended. (chapter 2, the academic achievements of 2)B, Based on space-time perception of service recommendation algorithm (TLCF)Paper caused by network dynamic change service QoS prediction researches on the problem of inaccurate.Consider impact on the quality of service time and space, will be introduced to the time and location information service QoS prediction, the user of the two-dimensional matrix is transformed into 3 d user-service-time matrix, service quality and service is closely related to the activation time, adjust the window size, can control the prediction precision.Considering the influence of the location information for service quality at the same time, to divide the IP address for service and user group, the same network segment of the user and the service is divided into the same group, makes the search service is restricted to small manageable.TLCF algorithm is put forward by the IP address as a close neighbor selection basis, predict the location of the correlation..(the third chapter, academic achievement (3)C, Based on SSN Web Services collaborative filtering recommendation algorithmIn cloud computing environment and demand for services recommended by reducing the need for service recommended search costs, service network SSN model is put forward.SSN model put forward by the distributed information storage and maintenance, to prevent the single point of failure caused by network dynamics, performance bottlenecks, and load imbalance problems.Mainly defines the relations between the two kinds of this model, the recommendation and trust relations, proposed to the recommendation and the calculation method of trust, the use of rewards and punishments attenuation factor and time factor to trust updated in real time, makes the closer the behavior for the influence of trust value, the greater the behavior characteristics of better reflect the current node.Collaborative filtering recommendation algorithm based on service network SSN-CF for can effectively prevent malicious behavior.The algorithm to improve the efficiency of search service recommendation accuracy has a practical and effective, (the fourth chapter, academic achievements of 4)D, Based on Service Network Service Composition Recommended algorithmFor recommendation of Service Composition, using ProgrammableWeb.com registered users, labels, Service and Service Composition (mashups) information, build a kind of by the user, tag, API (services), Mashup UMAT (services) constitute a Service network model, the model proposed to Service and Service combination of registration information directly applied to the Service and the Service portfolio recommendations, Based on Service network (-based on UMAT Service Composition Recommendation Algorithm UMAT-CSRA) recommended Service Composition algorithm, through various similarity of UMAT network computing, to set a Service request with UMAT network in a certain threshold value of similarity, to control through on the ProgrammableWeb.com web site the length of Service and Service portfolio recommendation Service.Through the real ProgrammableWeb.com registered users, tags, information services and service combination (mashups) experiment, from the indicators such as recall ratio and accuracy, UMAT-CSRA has good effect of service recommendation, and is a new way of thinking to solve the problem of service composition.