A Novel Rules Extraction Method Based on Clustering Analysis

Clustering is a method of unsupervised learning, and a common technique for statistical data analysis used in many fields, including machine learning, data mining, pattern recognition, image analysis and bioinformatics a novel algorithm based on clustering to extract rules from neural networks is proposed. After neural networks have been trained and pruned successfully, inner-rules are generated by discrete activation values of hidden units. According to discrete activation values of this hidden unit, cluster weights from input units to it. The incremental rules are extracted and the existing rule set is updated based on this algorithm. The result shows this method is quite valuable.