基于Bootstrap技术的Web服务QoS值的置信区间估计和预测

目前对Web服务QoS（Quality of Service）的预测研究，通常预测QoS的静态值，很少预测QoS值的置信区间.本文借助非参数统计学的Bootstrap技术，提出估计Web服务QoS值置信区间的方法；然后利用与当前Web用户相似的其他Web用户调用待预测Web服务的QoS历史数据，预测当前Web用户调用待预测Web服务的QoS值的置信区间.本文估计了WSDream数据集1中每个用户调用每个Web服务的QoS值的置信区间，实验发现这些置信区间的上下限近似服从重尾分布.通过随机选择WSDream数据集1中60%到90%的用户和Web服务作为训练集，预测另外10%到40%的用户和Web服务的QoS值，实验结果表明预测的QoS置信区间与估计的QoS置信区间的平均覆盖率超过70%，最高达76%.在服务选择或服务推荐时给用户提供一个估计的或预测的QoS置信区间，可以更好地满足用户的个性化需求.

An Approach to Estimate and Predict the Confidence Interval of Web Service QoS Based on Bootstrap

Nowdays we usually predict the static value of QoS (Quality of Service) rather than the confidence interval of the QoS in researches toward the prediction of Web services QoS.With the help of non-parametric statistical Bootstrap technique,we propose an approach to estimate and predict the confidence interval of Web services QoS; and then we use the historical QoS data of Web users which are similar to current Web users to predict the confidence interval of QoS values of the current Web users.Furthermore,we estimate the QoS confidence interval of each user invokes each Web service in WSDream dataset1.According to the experiment,we find out that the confidence interval follows a heavy tailed distribution.By randomly choosing 60% to 90% of users and services from WSDream dataset1 as our training dataset and predicting the QoS value of the other 10% to 40% users and services,we find that the average coverage rate is over 70% between the predicted QoS confidence interval and the estimated QoS confidence interval and the maximum average rate is as high as 76%.It is much better to meet personal requirement if we provide an estimated or predicted QoS confidence interval in the service selection or service recommendation.