An Accelerating Reduction Approach for Incomplete Decision Table Using Positive Approximation Set.

Area of Science:

• Data Science
• Artificial Intelligence
• Information Theory

Background:

• The exponential increase in sensor data necessitates efficient feature selection methods.
• Traditional feature selection algorithms struggle with large, incomplete datasets.
• Rough set theory offers a framework for analyzing imprecise and uncertain data.

Purpose of the Study:

• To develop a novel heuristic approach (ARIPA) for accelerating feature selection in incomplete decision tables.
• To enhance the efficiency and stability of existing attribute reduction algorithms using rough set theory.
• To investigate the application of ARIPA within classical and variable precision rough set models.

Main Methods:

• A new perspective on rough set theory using positive approximation sets and granularity domains.
• Development of the ARIPA heuristic approach for accelerating representative reduction.
• Implementation of ARIPA within classical and variable precision rough set models (ARIPA and ARIPA-IVPR).
• Integration of ARIPA to improve two state-of-the-art reduction algorithms.

Main Results:

• ARIPA significantly enhances the computational efficiency of attribute reduction algorithms.
• Experiments on UCI incomplete datasets demonstrate faster reduction task completion.
• Improved algorithms show comparable or superior stability to original methods.
• ARIPA-IVPR provides acceleration within the variable precision rough set model.

Conclusions:

• The proposed ARIPA approach effectively accelerates feature selection for incomplete decision tables.
• ARIPA offers a valuable enhancement for existing rough set-based reduction algorithms.
• The method demonstrates practical utility and improved performance on real-world datasets.