Accelerated Feature Selection via Discernibility Hashing: A Rough Set Approach.

Area of Science:

• Artificial Intelligence
• Data Science
• Information Theory

Background:

• The discernibility matrix is crucial for knowledge reduction in rough set theory.
• Existing algorithms face scalability issues due to the matrix's quadratic complexity with massive datasets.

Purpose of the Study:

• To develop a more efficient knowledge reduction method for rough set systems.
• To overcome the scalability limitations of traditional discernibility matrices.

Main Methods:

• Introduced a discernibility hashing strategy to manage attribute set growth.
• Mapped discernibility attributes to a one-dimensional hash space, reducing matrix dimensionality.
• Developed a feature selection algorithm utilizing the discernibility hash for efficient knowledge reduction.

Main Results:

• The discernibility hash method significantly reduces storage space compared to the traditional matrix.
• Experimental results demonstrate superior running times for the proposed algorithm.
• Invalid and redundant attribute sets are effectively eliminated from the reduction process.

Conclusions:

• Discernibility hashing offers a scalable and efficient solution for knowledge reduction in rough set theory.
• The proposed method enhances the practical applicability of rough set models to large-scale data.
• This approach represents a significant advancement in computational intelligence and data analysis.