Abstract: Aimed at the currently shortage of XML streams quality detecting,an adaptive quality forecasting method of XML data streams(XAFS) is presented.It uses the vector matrix built by the XML keys as windows,uses vector product wavelet transform to multilevel decompose and refactor the XML streams series,and then forecasts the XML streams’ quality combing squares support vector machines.For even more adapting net load,it proposes a control strategy by weight and adaptive adjustment to ensure XML streams quality.Contrasting with other corresponding algorithms,a series of emulation experiments shows that this method has proper rather precision and practicability,so it is an effective forecasting method of XML streams mining.