Estimating body weight in Sujiang pigs using artificial neural network, nearest neighbor, and CART algorithms: a...

Area of Science:

• Animal Science
• Machine Learning
• Genetics

Background:

• Accurate prediction of body weight (BW) in pigs is crucial for breeding programs and farm management.
• Morphological traits offer a non-invasive method for estimating BW.
• Evaluating machine learning algorithms for predictive accuracy in livestock is an ongoing area of research.

Purpose of the Study:

• To assess the efficacy of various machine learning algorithms in predicting Sujiang pig BW.
• To identify the most accurate and efficient algorithms for BW prediction using limited morphological data.
• To determine key morphological traits that best predict BW in Sujiang pigs.

Main Methods:

• Utilized morphological measurements (age, body length, backfat thickness, chest circumference, body height, chest width, hip width) from 365 mature Sujiang pigs.
• Applied Artificial Neural Network (ANN), K-nearest neighbors (KNN), and Classification and Regression Tree (CART) algorithms for BW prediction.
• Optimized ANN models using Levenberg-Marquardt, Bayesian regularization, and scaled conjugate gradient training algorithms.

Main Results:

• The ANN algorithm demonstrated superior performance, achieving an R² of 0.85.
• Hip width (HW), body length (BL), and body height (BH) were identified as significant predictors of BW.
• Excluding BH from the model reduced predictive accuracy by approximately 5%.

Conclusions:

• The ANN algorithm, particularly with Levenberg-Marquardt or Bayesian Regularization, is a powerful tool for accurate pig BW prediction.
• Identified morphological traits can serve as effective indirect selection criteria for improving breeding strategies.
• This study supports the use of machine learning for establishing selection criteria and breed standards in Sujiang pigs.