WSEAS Transactions on Computers
Print ISSN: 1109-2750, E-ISSN: 2224-2872
Volume 22, 2023
Optimization of Fuzzy Regression Transfer Learning using Genetic Algorithm for Cross-Domain Mapping
Author:
Abstract: Artificial intelligence and big data have become widely utilized in industry and thus machine learning has been extensively researched. However, it is challenging to apply existing data-driven methods when the amount of data is insufficient. Therefore, transfer learning, which reuses knowledge acquired from domains with similar data characteristics and tasks, has gained attention for achieving fast and accurate model learning in new domains. Although numerous transfer learning methods have been proposed for classification problems, few have been proposed for regression problems. Moreover, conventional fuzzy regression transfer learning tends to work well only in limited domain environments with extremely limited target data, making its application to real-world data challenging. The present study applies a combination of regression models based on Takagi-Sugeno fuzzy theory and transfers learning to regression problems in domains with incomplete knowledge. We propose two methods, one based on a genetic algorithm and one based on differential evolution combined with a genetic algorithm, for optimizing mapping for input space modification and applying them to real datasets. The results of evaluation experiments demonstrate that the proposed methods have higher efficiency and learning accuracy than those of conventional methods.
Search Articles
Keywords: Transfer learning, Fuzzy, Regression, Genetic algorithm, optimization, cross-domain mapping, input space modification, Differential Evolution
Pages: 316-323
DOI: 10.37394/23205.2023.22.36