ScholarLens Back to search
2002•Fundamenta InformaticaeRequires access

Approximate Entropy Reducts

Dominik Ślȩzak

Open publisher page 272 citations

Abstract

We use information entropy measure to extend the rough set based notion of a reduct. We introduce the Approximate Entropy Reduction Principle (AERP). It states that any simplification (reduction of attributes) in the decision model, which approximately preserves its conditional entropy (the measure of inconsistency of defining decision by conditional attributes) should be performed to decrease its prior entropy (the measure of the model's complexity). We show NP-hardness of optimization tasks concerning application of various modifications of AERP to data analysis.

About this research paper

What this paper is about

We use information entropy measure to extend the rough set based notion of a reduct. We introduce the Approximate Entropy Reduction Principle (AERP). It states that any simplification (reduction of attributes) in the decision model, which approximately preserves its conditional entropy (the measure of inconsistency of defining decision by conditional attributes) should be performed to decrease its prior entropy (the measure of the model's complexity). We show NP-hardness of optimization tasks concerning application of various modifications of AERP to data analysis.

Why it matters

OpenAlex reports 272 citations for this work. Citation counts describe recorded attention and do not establish research quality.

Key contribution

A contribution statement is not available in the OpenAlex record.

Method / approach

Method details are not available in the OpenAlex metadata.

Main findings

Findings are not separately available in the OpenAlex metadata.

Limitations

Limitations are not available in the OpenAlex metadata.

Applications

Application details are not available in the OpenAlex metadata.

Available abstract

We use information entropy measure to extend the rough set based notion of a reduct. We introduce the Approximate Entropy Reduction Principle (AERP). It states that any simplification (reduction of attributes) in the decision model, which approximately preserves its conditional entropy (the measure of inconsistency of defining decision by conditional attributes) should be performed to decrease its prior entropy (the measure of the model's complexity). We show NP-hardness of optimization tasks concerning application of various modifications of AERP to data analysis.

Key concepts: Mathematics, Computer science, Data mining

Related papers

Related paper 1Related paper 2Related paper 3
Back to paper searchBrowse research topicsOriginal source
Approximate Entropy Reducts — Research Paper | ScholarLens