Optimized support vector machines for unveiling mortality incidence in Tilapia fish

Abstract

In this paper, a new classification approach based on swarm-optimization is introduced to investigate the various effects of the ammonia concentration on the protein level and bioactivity that directly affect the Egyptian Nile’s fish health and mortality rate (i.e. Tilapia fish “O. Niloticus”). This approach enhances the Support Vector (SVM) Machines to classify the fish based on the protein level by Moth-Flame Optimization (MFO) algorithm. The experiment was divided into sub-phases: lab experiments and computational experiments. The primary purposes of the proposed approach, guiding decision-makers to review the pathophysiological status of the fish. The proposed MFO-SVM approach utilizes physical and chemical measurements to finally show revolutionary advances against the classic SVM and other well-known optimizers and classifiers. By achieving 99.983% of classification accuracy, the proposed approach outperforms other machine learning approaches on the same dataset. We believe that such an approach could be useful for many other real-world challenging tasks.

Keywords

• Support vector machine
• Tilapia fish protein
• Moth-flame optimization
• Fish mortality prediction