基于CART与SlopeOne的服务质量预测算法

摘要

针对现有的预测算法大多未有效利用用户一服务对的潜在特征问题,提出一种基于分类和SlopeOne的预测算法,通过用户一服务对的历史服务质量值提取出用户和服务的个性特征(用户和服务的服务质量均值与方差);基于提取出的特征,使用CART(classification and regression trees)对用户—服务对进行分类;使用SlopeOne算法在目标用户和目标服务所在的分类集合数据集上进行回归预测,提高了预测准确度;选用真实数据集WS-Dream进行实验,实验结果表明该方法在数据稀疏情况下具有较好的预测精度.%Aiming at the problem that the existing Web service QoS prediction approaches did not take the characteristics of users-service into consideration,a prediction algorithm based on Classification and Regression Trees (CART) and SlopeOne was proposed.The history QoS value was first employed to extract the users and services features and then classify user-service pair with CART based on the extracted features.To improve the prediction accuracy,SlopeOne was used to predict QoS on the target user and the target service's classification result.The experiments on a real-world dataset were conducted,and the experimental results illustrated that the proposed method had better prediction accuracy than baseline models.