Improving the genetic bee colony optimization algorithm for efficient gene selection in microarray data

• Published in: Progress in artificial intelligence 2018-12, Vol.7 (4), p.399-410
• Main Authors: Pino Angulo, Adrian, Shin, Kilho, Velázquez-Rodríguez, Camilo
• Format: Article
• Language: English
• Subjects: Algorithms; Ant colony optimization; Artificial Intelligence; Biomedical data; Computational Intelligence; Computer Imaging; Computer Science; Control; Critical components; Data analysis; Data mining; Data Mining and Knowledge Discovery; Filtration; Mechatronics; Natural Language Processing (NLP); Optimization algorithms; Pattern Recognition and Graphics; Regular Paper; Robotics; Vision; Workflow
• ISSN: 2192-6352; 2192-6360
• DOI: 10.1007/s13748-018-0161-9

Feature selection is a very critical component in the workflow of biomedical data mining applications. In particular, there is a need for feature selection methods that can find complex relationships among genes, yet computationally efficient. Within the scope of microarray data analysis, the genetic bee colony ( Gbc ) algorithm is one of the best feature selection algorithms, which leverages the combination between genetic and ant colony optimization algorithms to search for the optimal solution. In this paper, we analyse in depth the fundamentals lying behind the Gbc and propose some improvements in both efficiency and accuracy, so that researchers can even take more advantage of this excellent method. By (i) replacing the filtering phase of Gbc with a more efficient technique, (ii) improving the population generation in the artificial colony algorithm used in Gbc , and (iii) improving the exploitation method in Gbc , our experiments in microarray data sets reveal that our new method Gbc+ is not only significantly more accurate, but also around ten times faster on average than the original