An Attribute Reduction Method Using Neighborhood Entropy Measures in Neighborhood Rough Sets

Area of Science:

• Data Mining and Machine Learning
• Rough Set Theory
• Information Theory

Background:

• Classical rough set theory may lose information during discretization of continuous data.
• Neighborhood rough set theory offers an alternative for handling continuous data.
• Attribute reduction is crucial for efficient data mining and improving classification performance.

Purpose of the Study:

• To propose a novel attribute reduction method for neighborhood rough sets.
• To enhance classification performance on complex and continuous data.
• To integrate algebraic and informational views for attribute reduction.

Main Methods:

• Developed a new average neighborhood entropy measure combining neighborhood approximate precision and neighborhood entropy.
• Introduced decision neighborhood entropy to address uncertainty and noisiness in neighborhood decision systems.
• Derived properties and relationships of these entropy measures.
• Proposed a heuristic attribute reduction algorithm.

Main Results:

• The proposed method effectively handles continuous data while preserving classification information.
• New entropy measures provide a comprehensive analysis of knowledge content and uncertainty.
• Experimental results demonstrate significant improvements in classification performance.
• The method successfully identifies the most relevant attributes.

Conclusions:

• The novel attribute reduction method based on neighborhood entropy measures is effective for complex data.
• The integration of algebraic and informational views enhances attribute reduction capabilities.
• The proposed approach offers a robust solution for data mining preprocessing with continuous data.