Final weight prediction from body measurements in Kıvırcık lambs using data mining algorithms

Area of Science:

• Animal Science
• Machine Learning
• Agricultural Engineering

Background:

• Accurate estimation of lamb live weight is crucial for livestock management and economic evaluation.
• Traditional methods may be time-consuming and labor-intensive.
• Exploring advanced computational methods can improve weight estimation efficiency.

Purpose of the Study:

• To evaluate the performance of various machine learning algorithms for estimating the final live weight of Kıvırcık lambs.
• To identify the most effective algorithm for predicting lamb weight based on morphological measurements.
• To assess the utility of body measurements in developing accurate weight estimation models.

Main Methods:

• Collected morphological data including height at withers, back height, croup height, chest depth, body length, chest width, and chest circumference from Kıvırcık lambs.
• Applied and compared six machine learning algorithms: Chi-square automatic interaction detection (CHAID), exhaustive CHAID, Classification and Regression Tree (CART), Random Forest (RF), Multivariate Adaptive Regression Spline (MARS), and Bootstrap-aggregating MARS (Bagging MARS).
• Evaluated algorithm performance using goodness-of-fit criteria such as R-squared, Adjusted R-squared, Coefficient of Variation (CV), Mean Square Error (MSE), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE).

Main Results:

• The Multivariate Adaptive Regression Spline (MARS) algorithm demonstrated the highest predictive accuracy, achieving the best goodness-of-fit.
• MARS model performance was optimized using cross-validation and parameter optimization.
• The study found significant correlations between morphological features and lamb live weight, enabling successful weight estimation.

Conclusions:

• Morphological measurements of Kıvırcık lambs can be effectively utilized with machine learning algorithms for accurate live weight estimation.
• The MARS algorithm is a highly suitable tool for developing robust models for predicting lamb live weight.
• This approach offers a promising, data-driven method for enhancing livestock management practices.