Research And Design Of Context-aware Web Service Recommendation Method

Both the functional attributes and non-functional attributes are considered when recommending web services. In mobile environment,when invoking the web service, the situation of the users is not always the same. The existing context-aware web service recommendation methods are lack of specific analysis and selection of context properties, which are mostly based on common sense to choose context properties for service recommendation.In order to solve these problems, this paper proposes a context-aware web service recommendation method based on analysis of the relevance between context properties and QoS. This method puts forward the calculation formula of correlation between context properties and QoS to select context properties that are related to the QoS. Then,both the context-aware method and the collaborative filtering method are used to predict the QoS to recommend the optimal web service and the accuracy of prediction is improved. This paper mainly completes the following research work:1) According to the disadvantages of traditional algorithms, we propose a new method of context properties and QoS relevance calculation.The data of context properties and QoS is obtained from user history, by clustering the data, the correlation between context properties and QoS is transformed into the correlation between computational vectors. Combined with the LAN distance, this paper optimizes the correlation calculation, and solves the correlation calculation error when some vector data is very small.2) We propose a context-aware web service recommendation method based on analysis of the relevance between context properties and QoS.This method calculates the correlation between the context properties and QoS and select the associated context properties based on Correlation threshold. Those related context properties are exploited to filter the original data for finding users whose context is similar to the current user.Then users whose QoS experience is similar to the current user are found too. The data of the forecast of the uses whose context is similar to the current user and the users whose QoS experience is similar to the current user are exploited to forecast QoS for service recommendation.3) Real data is used to conduct simulation experiments to verify that selecting related context properties can improve the accuracy of service recommendation. Experiments show that compared with other service recommendation methods, the proposed method has higher recommendation accuracy. The experiment also analyzes the impact of correlation threshold of the proposed method.