Discrete Homogeneous and Non-Homogeneous Markov Chains Enhance Predictive Modelling for Dairy Cow Diseases

Area of Science:

• Veterinary epidemiology
• Agricultural technology
• Data science

Background:

• Dairy cow disease prediction is crucial for herd health management and economic viability.
• Existing machine learning models face limitations due to disease-specific development and scarce farm data.

Purpose of the Study:

• To develop a novel predictive model for dairy cow diseases using Markov chains.
• To address data limitations inherent in machine learning approaches for disease prediction.

Main Methods:

• Utilized discrete Homogeneous and Non-homogeneous Markov chains for disease modeling.
• Developed a method for determining the optimal number of Markov chain states.
• Employed Chebyshev distance minimization for selecting the best predictive model.

Main Results:

• Achieved less than 15% maximum difference between actual and predicted data for 14 out of 19 diseases.
• Demonstrated a model adaptable for low-tech dairy farm implementation.
• Showcased potential for extension to other disease types with minimal adjustments.

Conclusions:

• The Markov chain model provides an effective solution for dairy cow disease prediction, overcoming data scarcity issues.
• This model can enhance decision support systems, leading to improved herd health and evidence-based farming strategies.
• Facilitates cost projection for treatments like antibiotics in dairy farming.