Feature selection strategy based on hybrid crow search optimization algorithm integrated with chaos theory and fuzzy c-means algorithm for medical diagnosis problems

• Published in: Soft computing (Berlin, Germany) Vol. 24; no. 3; pp. 1565 - 1584
• Main Authors: Anter, Ahmed M., Ali, Mumtaz
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.02.2020; Springer Nature B.V
• Subjects: Accuracy; Artificial Intelligence; Chaos theory; Computational Intelligence; Control; Cost function; Datasets; Diagnosis; Engineering; Exploitation; Feature selection; Food; Fuzzy logic; Fuzzy sets; Global optimization; Knowledge acquisition; Liver; Local optimization; Machine learning; Mathematical Logic and Foundations; Mechatronics; Medical diagnosis; Methodologies and Application; Optimization algorithms; Random variables; Robotics; Search algorithms; Time series; Feature selection; Crow search optimization; Medical diagnosis; Chaos theory; Fuzzy c-means
• ISSN: 1432-7643; 1433-7479
• DOI: 10.1007/s00500-019-03988-3

Powerful knowledge acquisition tools and techniques have the ability to increase both the quality and the quantity of knowledge-based systems for real-world problems. In this paper, we designed a hybrid crow search optimization algorithm integrated with chaos theory and fuzzy c-means algorithm denoted as CFCSA for feature selection problems of medical diagnosis. In the proposed CFCSA framework, the crow search algorithm adopts the global optimization technique to avoid the sensitivity of local optimization. The fuzzy c-means (FCM) objective function is used as a cost function for the chaotic crow search optimization algorithm. The proposed algorithm CFCSA is benchmarked against the binary crow search algorithm (BCSA), chaotic ant lion optimization algorithm (CALO), binary ant lion optimization algorithm (BALO) and bat algorithm relevant methods. The proposed CFCSA algorithm vs. BCSA, CALO, BALO and bat algorithm is tested on diabetes, heart, Radiopaedia CT liver, breast cancer, lung cancer, cardiotocography, ILPD, liver disorders, hepatitis and arrhythmia. Experimental results show the proposed method CFCSA is better against comparative models in feature selection on the medical diagnosis data sets.