Abstract: In developing algorithms for data stream mining, little work has been done in the management of various algorithms. Although there have been many effective algorithms developed specifically for data stream mining, the application rate is quite low. In this research, we establish a process model for data stream mining. A process model for the algorithm management framework based on web services, PMAMF-DSM, is proposed. We analyze the construction of data stream algorithm repository and the architecture of the framework. Using the framework, a data stream oriented algorithm management system is implemented on Eclipse. Experiments indicate that the framework has high flexibility and self adaptability.

Key words: data mining, process model, data streams, predictive model markup language (PMML), web service