Attribute reduction based on classes in incomplete ordered decision systems.

Area of Science:

• Data Science
• Artificial Intelligence
• Machine Learning

Background:

• Dominance-based rough set approach (DRSA) is widely used for multi-criteria decision-making with preference-ordered data.
• Real-world data often presents challenges due to incomplete information, necessitating robust attribute reduction techniques.
• Traditional attribute evaluation methods in incomplete ordered decision systems (IODS) can be computationally expensive and may overlook classification impact.

Purpose of the Study:

• To propose novel attribute reduction approaches for incomplete ordered decision systems (IODS).
• To develop a new criterion for attribute importance evaluation that addresses limitations of existing methods.
• To enhance the efficiency and classification accuracy of attribute reduction algorithms.

Main Methods:

• Defined inter-class proximity and whole inter-class non-dominance concepts within an IODS.
• Developed a method for calculating inter-class non-dominance using an expanded dominance matrix.
• Designed the Heuristic Attribute reduction based on inter-class Non-DOminance (HANDR) algorithm, incorporating a heuristic search strategy.

Main Results:

• The proposed HANDR algorithm demonstrates significant improvements in running time compared to traditional methods.
• Experiments show a notable increase in classification accuracy using the HANDR algorithm on UCI datasets.
• Whole inter-class non-dominance proves to be an effective criterion for attribute importance in IODS.

Conclusions:

• The HANDR algorithm offers an efficient and accurate solution for attribute reduction in incomplete ordered decision systems.
• The proposed evaluation criterion, whole inter-class non-dominance, effectively addresses the challenges of incomplete information in decision-making.
• This research contributes to advancing multi-criteria decision-making techniques for complex, real-world datasets.