Granular computing: An augmented scheme of degranulation through a modified partition matrix

As an important technology in artificial intelligence, Granular Computing has emerged as a new multi-disciplinary paradigm and received much attention recent years. Information granules forming abstract efficient characterization of large volumes numeric data have been considered the fundamental constructs Computing. By generating centroids (prototypes) partition matrix, fuzzy clustering is commonly encountered way information granulation. reverse process granulation, degranulation involves reconstruction completed on basis granular representatives (decoding into data). Previous studies shown that there relationship between error performance granulation process. Typically, lower is, better becomes. However, existing methods usually cannot restore original data, which one reasons behind occurrence error. To enhance quality (degranulation), this study, we develop augmented scheme through modifying matrix. proposing scheme, elaborate novel collection granulation-degranulation mechanisms. In constructed approach, prototypes can be expressed product dataset matrix Then, process, reconstructed decomposed prototypes. series operations. We offer thorough analysis developed scheme. The experimental results are agreement with underlying conceptual framework. obtained both synthetic publicly available datasets reported to show enhancement thanks proposed method. It pointed out by using approach some cases errors reduced close zero approach.