Abstract: In the analysis of collaborative prediction approaches with Pearson Correlation Coefficient (PCC),the significant ratio relation was found between Web services and Quality of Service(QoS),while QoS prediction error  might caused by  ignoring the relationship of this ratio.To deal with the problem,a prediction approach considering the similarity ratio SRPre was proposed.Based on the historical QoS data,the combination of user-based approach and item-based approach was used.A minimum common call number threshold was introduced to reduce the calculation error of similarity,and a minimum similarity threshold was used to screen the similar neighbors.According to QoS average value ration of predicted object to similar neighbors,the similarity ratio was computed.The similarity ratio was added to the process of QoS collaborative prediction.The experimental results showed that the proposed approach was effective in improving the prediction accuracy.

Key words: Web service, quality of service prediction, collaborative filtering, Pearson correlation coefficient

摘要： 针对基于皮尔逊相关系数计算相似度的协同预测方法中,忽略Web服务QoS之间的比率关系而造成的预测误差问题,提出一种考虑相似比率的Web服务QoS协同预测方法SRPre。该方法基于历史QoS数据,将基于用户和基于服务的协同预测方法相融合,引入最小共同调用数阈值削减相似度计算误差,利用最小相似度阈值筛选相似邻居,根据预测对象QoS与相似邻居QoS平均值之比计算相似比率,在协同预测计算过程中加入相似比率获得QoS预测值,通过实际Web服务实验说明了所提该方法在提高预测准确度方面的有效性。

关键词: Web服务, QoS预测, 协同过滤, 皮尔逊相关系数