Year: 2024
Type: article
Abstract: This paper presents a novel method for feature selection in binary classification tasks based on histogram-based scoring. By leveraging the distribution differences between feature values associated with positive and negative classes, we generate a score to determine the most informative features. The method, called Histogram-Based Feature Selection (HBFS), has been tested against a variety of datasets and compared to the Fisher Score for performance assessment. Our findings indicate that HBFS either matches or outperforms Fisher Score in most datasets.
Source: DergiPark (Istanbul University)
Authors: Delil, Selman, Ağraz, Melih, Kuyumcu, Birol
Institutions: Istanbul Technical University, Giresun University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4405708086