Loading…

Intuitionistic Fuzzy Quantifier and Its Application in Feature Selection

Nowadays, databases expand rapidly due to electronically generated information from different fields like bioinformatics, census data, social media, business transactions, etc. Hence, feature selection/attribute reduction in databases is necessary in order to reduce time, cost, storage and noise for...

Full description

Saved in:
Bibliographic Details
Published in:International journal of fuzzy systems 2019-03, Vol.21 (2), p.441-453
Main Authors: Singh, Shivani, Shreevastava, Shivam, Som, Tanmoy, Jain, Pankhuri
Format: Article
Language:English
Subjects:
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Nowadays, databases expand rapidly due to electronically generated information from different fields like bioinformatics, census data, social media, business transactions, etc. Hence, feature selection/attribute reduction in databases is necessary in order to reduce time, cost, storage and noise for better accuracy. For this purpose, the rough set theory has been played a very significant role, but this theory is inefficient in case of real-valued data set due to information loss through discretization process. Hybridization of rough set with intuitionistic fuzzy set successfully dealt with this issue, but it may radically change the outcome of the approximations by adding or ignoring a single element. To handle this situation, we reconsider the hybridization process by introducing intuitionistic fuzzy quantifiers into the idea of upper and lower approximations. Supremacy of intuitionistic fuzzy quantifier over VPRS and VQRS is presented with the help of some examples. A novel process for feature selection is given by using the degree of dependency approach with intuitionistic fuzzy quantifier-based lower approximation. A greedy algorithm along with two supportive examples is presented in order to demonstrate the proposed approach. Finally, proposed algorithm is implemented on some benchmark datasets and classification accuracies for different classifiers are compared.
ISSN:1562-2479
2199-3211
DOI:10.1007/s40815-018-00603-9