The Cuts Selection Method Based on Histogram Segmentation and Impact on Discretization Algorithms

Area of Science:

• Data Science
• Machine Learning
• Rough Set Theory

Background:

• Data preprocessing, including discretization, is crucial for rough set theory and entropy-based methods.
• Understanding the relationship between data histogram segmentation and data discretization is essential for improving data quality.

Purpose of the Study:

• To propose and evaluate a novel data segmentation technique based on histograms for enhancing data discretization quality.
• To investigate the connection between histogram segmentation, data reducts, and classification rule generation.

Main Methods:

• A new data segmentation technique using histogram analysis was developed.
• The significance of cut positions and their relation to data reducts were examined.
• The proposed Cuts Selection Method was integrated with the Maximal Discernibility (MD) algorithm and entropy-based discretization.

Main Results:

• Reduct attributes exhibit more irregular histograms compared to non-reduct attributes.
• Histogram segmentation effectively guides the selection of cuts, improving discretization quality.
• Classification results using the proposed method showed improvements when compared to the Naïve Bayes algorithm.

Conclusions:

• Histogram-based data segmentation is a beneficial approach for data discretization in rough set theory.
• The proposed Cuts Selection Method offers an effective way to enhance data preprocessing for classification tasks.