International Conference on Instrumentation, Measurement, Circuits and Systems (ICIMCS 2011)
267 Research of Outlier Mining Framework for Data Streams Based on Multiagent System and its Application
Download citation file:
- Ris (Zotero)
- Reference Manager
In order to generalize the outlier data mining to large-scale data streams, a distributed mining framework is proposed in this paper. The proposed framework is constructed by Multi-Agent system. It is composed of data window agent, data balanced agent, local outlier data mining agent and global outlier data selection agent. The data streams are separated by data window agent first, and find the local outlier data by corresponding agent. If necessary, the data on each distributed node will be balanced by the data balanced agent. At last, the global outlier data are selected by a parallel structure. The framework proposed in this paper has been applied in City Operation System and testified by practice well and feasible.