Text Classification Based on Rule Mining by Granule Network Constructing

Text classification is one of the practices of knowledge discovery. Designation of the classifier is the most important par of text classification. Comparing with the methods based on statistic theory, classification based on rule learning is a better one on some situations. A granular computing approach is proposed to learn rules by constructing a granule network while classifying texts. The algorithm of constructing granule network is involved in a refining process that pick-up second-layer granules from the largest granule, third-layer granules from second-layer ones, repeat this procedure until get the smallest granules. During the work, the whole granule network is completed and text classification rules are learned.