Extracting Rules and Knowledge in Multi-Level Information Table

Abstract

Rules extraction is a basic issue in both knowledge representation and data mining. In this paper, we put forward an approach to rules extraction by considering the regular condition entropy and the mutual information in a multi-level information table. The multi-level information table is investigated by introducing an real life example. Then the attribute value conversion function is constructed in the multi-level information table to obtain the higher levels attribute values from the lower levels. Moreover, the thickness degree relationships between different global levels is presented in detail. Finally, an example on rule extraction from some commodities is applied and tested to illustrate the effectiveness and rationality of our method.