Abstract: Due to the different backgrounds and experiences, users often have different impressions of the same QoS though they have been served all the same. An approach to predict QoS was therefore proposed, based on not only the historical QoS data, but also the factors of environment and users' inputs. Normalizing the original data by Gauss method, a set of Web services highly similar to the target service was picked out via the similarity computation to compare the target service to be predicted with other services. Then, the blank datasheet was predicted according to the service similarity so as to compute the similarity between users for the pridiction of QoS data of users' target service. Experimental results showed that the approach proposed can improve the prediction accuracy of QoS for Web service.