Leveraging Chicken Swarm Algorithm for Feature Selection Optimization Targeting Efficient Patients Backlog Elimination

Abstract

The exponential growth of data sources brings the challenge of maintaining the processing performance and reducing computation complexity. One of the vital solutions is the success in preserving the significant attributes. Consequently, this research focuses on proposing an effective method for feature selection which is based on the adaptation of the chicken swarm optimization algorithm. The research focuses on adapting the algorithm strategy in the process of searching the data space from random-based strategy to a more systematic method which ensures raising the algorithm performance. The study proposes a novel search strategy based on applying an effective clustering technique to effectively identify the main algorithm players which consequently enhance the algorithm performance. On the other hand, focusing on business objectives, this research proposes a novel framework that focuses on the patients’ backlogs. The study applies the proposed enhancement to eliminate the patients’ backlog while maintaining the prioritization. The proposed framework is generic and could be applied to the concept of backlogs in any domain. The experiment succeeded in confirming the applicability of the proposed adaptation for the chicken swarm optimization algorithm and reaching the business goal with a minimum accuracy percentage equal to 95.3% for random forest and a maximum of 98.9 for naive Bayes