Abstract: The accurate prediction of QoS is an important criterion for the Web services．In traditional QoS prediction method,when the time-weighted average and simple weight of various parameters are used,but a large number of Web services can not be accurately predicted,and the outcome is vague．An accurate prediction method of QoS based on global subspace decomposition mining was proposed,and the idea of global analysis of all data preprocessing was done,on this basis,through subspace methods,the subspace decomposition analysis of the data was carried out and the depth features of the data were extracted,then the global result of the analysis and subspace decomposition analysis results were fuzzed for effective data fusion and achieving the analytical data valid mining and predictive．A set of Web nodes and metrics were used to do the prediction experiment,and the result shows that with the global subspace decomposition mining method,the gradual process can be predicted with detailed process.The result is accurate and it has wide application value in QoS prediction.

Key words: QoS prediction,Global analysis,Subspace decomposition,Data mining