Evolving kernel extreme learning machine for medical diagnosis via a disperse foraging sine cosine algorithm

Area of Science:

• Computational intelligence
• Machine learning
• Medical informatics

Background:

• Kernel Extreme Learning Machine (KELM) is a popular classification and identification tool.
• KELM model performance is highly dependent on parameter optimization.
• Effective parameter optimization is crucial for practical KELM applications.

Purpose of the Study:

• To propose a novel parameter optimization strategy for KELM.
• To enhance the exploration and exploitation capabilities of optimization algorithms.
• To develop and validate a new machine learning model for medical diagnosis.

Main Methods:

• A Disperse Foraging Sine Cosine Algorithm (DFSCA) was developed to improve search agent exploration.
• DFSCA was integrated with KELM, creating the DFSCA-KELM model.
• The DFSCA's capabilities were tested using the CEC2017 benchmark suite.
• DFSCA-KELM was evaluated on six UCI medical datasets for diagnostic accuracy.
• The model was applied to two practical medical case studies.

Main Results:

• DFSCA demonstrated superior exploration and exploitation capabilities.
• DFSCA-KELM achieved effective performance in medical diagnosis tasks across multiple datasets.
• The model proved capable of handling real-world medical problems effectively.

Conclusions:

• The proposed DFSCA parameter optimization strategy significantly enhances KELM performance.
• DFSCA-KELM is a robust and effective tool for medical diagnosis.
• The developed technique shows promise for practical applications in the medical field.