Development of fast algorithms for reduct computation
Date
2022-12
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Benemérita Universidad Autónoma de Puebla
Abstract
"Rough Set Theory deals with imperfect knowledge in machine learning. Within Rough Set Theory, reducts are minimal subsets of attributes that preserve the discernibility power of the complete set of attributes in a dataset. Reducts are specially useful as an attribute reduction technique for classification and data storage. Unfortunately, computing all reducts of a dataset has exponential complexity regarding the number of attributes. Therefore, we proposed here a hardware approach for computing all reducts. The proposed platform outperforms previous hardware and software implementations in terms of runtime. In addition, a new algorithm for computing all reducts that uses simple operations for candidate evaluation, which is the fastest algorithm for an specific kind of datasets, was introduced. Furthermore, an experimental study for finding a relation between some properties of a dataset and the fastest algorithms for computing reducts, is presented. This study provides a guide for determining the appropriated algorithm for a specific problem. We proposed finally a new algorithm for computing all the shortest reducts. This new algorithm outperforms all other state–of–the–art algorithms for most datasets".
Description
Keywords
Citation
Collections
Document Viewer
Select a file to preview:
Can't see the file? Try reloading