Federated Multi-Label Learning (FMLL): Innovative Method for Classification Tasks in Animal Science.

Area of Science:

• Machine Learning
• Data Science
• Animal Science

Background:

• Federated learning facilitates collaborative model training without data sharing.
• Multi-label learning addresses classification where instances belong to multiple classes simultaneously.
• Combining these fields addresses privacy-preserving multi-label classification challenges.

Purpose of the Study:

• Introduce Federated Multi-Label Learning (FMLL) by integrating federated learning and multi-label classification.
• Apply the Binary Relevance (BR) strategy with Reduced-Error Pruning Tree (REPTree) for multi-label data.
• Evaluate FMLL's performance on diverse animal science datasets.

Main Methods:

• Developed a Federated Multi-Label Learning (FMLL) framework.
• Utilized the Binary Relevance (BR) strategy for multi-label data decomposition.
• Employed Reduced-Error Pruning Tree (REPTree) as the base classifier.
• Conducted experiments on Amphibians, Anuran-Calls-(MFCCs), and HackerEarth-Adopt-A-Buddy datasets.

Main Results:

• Achieved accuracy rates of 73.24% (Amphibians), 94.50% (Anuran-Calls), and 86.12% (HackerEarth-Adopt-A-Buddy).
• Demonstrated significant improvements exceeding 10% in average accuracy, precision, recall, and F-score compared to state-of-the-art methods.
• Validated FMLL's effectiveness across varied animal science datasets.

Conclusions:

• FMLL offers a robust and privacy-preserving solution for multi-label classification tasks.
• The proposed method shows superior performance over existing approaches in relevant benchmarks.
• FMLL has strong potential for applications in sensitive data domains like animal science.