HRFSVM: identification of fish disease using hybrid Random Forest and Support Vector Machine

Area of Science:

• Aquaculture
• Fish Health Management
• Machine Learning Applications

Background:

• Aquaculture fish diseases threaten global food security.
• Difficulty in distinguishing similar fish species complicates early disease detection.
• Lack of infrastructure hinders timely identification of infected fish in aquaculture.

Purpose of the Study:

• To propose a machine learning technique for identifying and categorizing fish diseases.
• To develop a Deep Convolutional Neural Network (DCNN) based method.
• To enhance disease detection accuracy in aquaculture.

Main Methods:

• A novel hybrid algorithm, Whale Optimization Algorithm with Genetic Algorithm (WOA-GA) and Ant Colony Optimization, was developed for global optimization.
• A hybrid Random Forest algorithm was employed for classification.
• A WOA-GA-based DCNN architecture was designed and compared with existing machine learning methods.

Main Results:

• The proposed WOA-GA-based DCNN architecture demonstrated improved quality and distinctions compared to current methods.
• Performance metrics including sensitivity, specificity, accuracy, precision, recall, F-measure, NPV, FPR, FNR, and MCC were evaluated.
• Effectiveness was validated using MATLAB.

Conclusions:

• The developed machine learning technique shows promise for early and accurate identification of fish diseases in aquaculture.
• The hybrid WOA-GA optimization combined with DCNN offers a robust approach to disease classification.
• This advancement can significantly contribute to safeguarding aquaculture productivity and food supply.