SM-RuleMiner: Spider monkey based rule miner using novel fitness function for diabetes classification

Area of Science:

• Medical Informatics
• Computational Intelligence
• Data Mining

Background:

• Diabetes mellitus presents a significant global health burden.
• Current rule-based systems for diabetes diagnosis face challenges in optimizing rulesets for accuracy, sensitivity, and specificity.
• Existing methods struggle to balance these critical performance metrics effectively.

Purpose of the Study:

• To introduce a novel Spider Monkey Optimization-based rule miner (SM-RuleMiner) for enhanced diabetes classification.
• To develop a new fitness function for SM-RuleMiner to achieve a comprehensive, optimal ruleset balancing accuracy, sensitivity, and specificity.
• To evaluate the performance of SM-RuleMiner against established and meta-heuristic-based rule mining algorithms.

Main Methods:

• Development and implementation of the Spider Monkey Optimization-based rule miner (SM-RuleMiner).
• Incorporation of a novel fitness function within SM-RuleMiner for balanced optimization.
• Comparative analysis using the Pima Indians Diabetes dataset with 10-fold cross-validation.
• Benchmarking against ID3, C4.5, CART, and other meta-heuristic algorithms.

Main Results:

• The proposed SM-RuleMiner demonstrated superior average classification accuracy and average sensitivity compared to benchmark algorithms.
• SM-RuleMiner achieved a smaller mean rule length and mean ruleset size, indicating greater efficiency.
• The novel fitness function effectively balanced accuracy, sensitivity, and specificity in the generated ruleset.

Conclusions:

• SM-RuleMiner offers a promising advancement in rule-based diabetes classification, surpassing traditional and meta-heuristic methods.
• The algorithm's ability to generate concise and accurate rulesets addresses key limitations of existing systems.
• This approach holds potential for improving the efficiency and effectiveness of diabetes diagnosis systems.