| A web service is actually an application built on top of a network, it can be described, published, found and invoked under web environment. Although the web services provide a convenience to integrate applications, they are often designed to be simple. What if no any single web service is located for a user request? The most straightforward way should be to put a finite set of services together such that any service can not fulfill the user request solely, but they can provide an assembled functionality which is sufficient for the user. Due to the increasing number of available services, each task in a composite service can be performed by a large amount of services that offer the same functionality. Then, how to choose for each task the most appropriate service from a set of candidates is important. Though the candidates for a task have the same functionality, they can be differentiated and selected by their non-functional properties (QoS-quality of service). But a single consumer cannot have sufficient pre-acquired data for all candidate web services. Most of the web services deployed on the Internet do not permit testing, and some services require consumers to pay. Therefore, QoS prediction of web services plays an important role in selecting services when a consumer wants to try the services which he never used.This thesis mainly discusses QoS prediction and QoS-aware Web service composition, its main contents and contributions are as follows:(1) The advantages and disadvantages of the existing QoS prediction method are analyzed and a QoS prediction method based on collaborative filtering is proposed. This method overcomes the disadvantages of traditional QoS prediction method (arithmetic mean) do not take the user's preferences and experience into account. The basic idea of our approach is to find out similarity between the consumers with collected QoS data and then to make prediction for the unused services based on the similarity. In order to overcome some limitations of the most widely used similarity mining method, we use the PEP similarity mining method in this thesis. This similarity mining method computes the similarity of consumers from three different views: proximity, impact and popularity.(2) The type of the QoS data is classified and a QoS data normalization method based on fuzzy set and membership functions is introduced. For accurate QoS data, in order to avoid extremely large or small values disturb the result of normalization, we use Gaussian approach to map these values out of the range of [0,1] to 0 or 1. For fuzzy QoS data, we find their membership functions and use their membership functions to map these values on the range of [0,1].(3) The advantages and disadvantages of the existing QoS awared web service composition method are analyzed and a QoS-awared web service composition method based on improved Genetic Algorithm (GA) is proposed. We use a modified GA to suit the requirement of the real-time web service composition, that is, use ant algorithm to select a population of antibodies as the first step of GA instead of generating initial population of antibodies randomly. In order to get a better convergence speed and speed up the pace of find the optimal solution, each available candidate services is encoded into a binary string to reduce the search space. |
PDF Full Text Request