INTERVAL ANALYSIS-BASED HYPERBOX GRANULAR COMPUTING CLASSIFICATION ALGORITHMS

Representation of a granule, relation and operation between two granules are mainly researched in granular computing. Hyperbox granular computing classification algorithms (HBGrC) are proposed based on interval analysis. Firstly, a granule is represented as the hyperbox which is the Cartesian product of $N$ intervals for classification in the $N$-dimensional space. Secondly, the relation between two hyperbox granules is measured by the novel positive valuation function induced by the two endpoints of an interval, where the operations between two hyperbox granules are designed so as to include granules with different granularity. Thirdly, hyperbox granular computing classification algorithms are designed on the basis of the operations between two hyperbox granules, the fuzzy inclusion relation between two hyperbox granules, and the granularity threshold. We demonstrate the superior performance of the proposed algorithms compared with the traditional classification algorithms, such as, Random Forest (RF), Support Vector Machines (SVMs), and Multilayer Perceptron (MLP).