A Multi-Level Granular Classification Model Based on Granularity Refinement

A multi-level granular classification model (MGCM) based on granularity refinement is proposed in this paper. First, based on the principle of justifiable granularity, a series of information granules are constructed. Then, the information granules are refined with different granularity levels according to the uncertainty of the information granules. Accordingly, the granular classification model is composed of serval sub-classification models with different granularity levels. The complexity and classification accuracy of the model can be taken into account when the data are described by information granules with different granularity levels. In each sub-model, the fuzzy C-means (FCM) method is considered to classify data. And particle swarm optimization algorithm is used to optimize the parameters. Some numerical examples are provided to illustrate the validity of the proposed model.