Developing a Robust Fuzzy Inference Algorithm in a Dog Disease Pre-Diagnosis System for Casual Owners

Area of Science:

• Veterinary Medicine
• Artificial Intelligence
• Machine Learning

Background:

• The growing pet market in Korea presents challenges for owners managing pet health.
• Non-expert pet owners often struggle with timely diagnosis of their pet's health issues.

Purpose of the Study:

• To develop a pre-diagnosis system for pet dogs using neuro-fuzzy learning.
• To enable non-expert users to monitor pet health by inputting observed symptoms.
• To provide likely disease predictions and coping strategies.

Main Methods:

• A disease-symptom database was created with veterinary guidance.
• Evaluation of three fuzzy inference algorithms: PFCM-R, FHAL, and the proposed MNFL.
• Development of the multi-layered neuro-fuzzy learner (MNFL) for improved robustness.

Main Results:

• PFCM-R showed high accuracy with clean data but poor noise tolerance.
• FHAL offered better noise tolerance but lower precision.
• MNFL achieved 98% accuracy even with noisy inputs, demonstrating superior robustness.

Conclusions:

• The MNFL system effectively aids in early detection of pet health issues.
• This empowers pet owners for better care and informed veterinary consultations.
• The system enhances overall companion animal well-being through improved health monitoring.