Abstract:

A modified knowledge discovery algorithm is proposed in this paper. In view of the limitations of the original pattern discovery algorithm based on residual analysis and recursive partitioning, different discretization criteria are adopted during the partitioning of sample space because different attributes have different contributions to pattern discovery. The proposed algorithm automatically adjusts the pattern-classifying criterion according to the subspace number. The application results on synthetic data set and power system stability assessment show that the proposed algorithm is rational and effective, and that it is of higher knowledge discovery efficiency and wider application, as compared with the original algorithm.

Key words: knowledge discovery, pattern discovery, diseretization, power gird, security assessment