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

The objectives of this study were to evaluate different machine learning algorithms for predicting body weight (BW) in Sujiang pigs using the following morphological traits: age, body length (BL), backfat thickness (BFT), chest circumference (CC), body height (BH), chest width (CW), and hip width (HW). Additionally, this study also investigated which machine learning algorithms could accurately and efficiently predict body weight in pigs using a limited set of morphological traits. For this purpose, morphological measurements of 365 mature (180 ± 5 days) Sujiang pigs from the Jiangsu Sujiang Pig Breeding Farm in Taizhou, Jiangsu Province, China were used. The age of the pigs (180 ± 5 days) was also included as a nominal predictor. In total, 218 individual measurements were obtained after data preprocessing. In the Sujiang pig dataset, BW had a significantly positive and high linear relationship with BH, BL, CW, HW, and CC resulting in values of 0.66, 0.72, 0.81, 0.84, and 0.88, respectively (p < 0.01). Artificial neural network (ANN), K-nearest neighbors (KNN), and classification and regression tree (CART) algorithms were used to predict BW. Overall, the ANN algorithm outperformed the other algorithms in this pig dataset according to the goodness of fit criteria of R² = 0.85, RMSE = 3.98, MAD = 3.25, MAPE = 4.25, SDR = 0.39, RAE = 0.002, MRAE = 0.008, and AIC = 97.96. The ANN algorithm was trained using several training algorithms, such as the Levenberg‒Marquardt algorithm, Bayesian regularization, and scaled conjugate gradient. In addition, the number of neurons in the hidden layer was manipulated to 2, 3, or 4. All training algorithms yielded similar results. However, when the predictor variables were HW, BL, and BH, the Levenberg-Marquardt network had the best ability to predict body weight in Sujiang pigs (R² = 0.83). When BH measurements were not included in the model, the model's predictive ability decreased by approximately 5%. According to the results, the use of Levenberg‒Marquardt and Bayesian Regularization in the ANN algorithm can help improve breeding strategies. The traits determined to be the best predictors of BW in Sujiang pigs using the ANN algorithm can be used as indirect selection criteria in the future. This study suggests that different age stages, breeds, and morphological traits can be used to accurately predict BW in pigs in future research. These findings indicate that the ANN algorithm is a powerful tool for accurately predicting pig BW using a limited set of traits. The results of the ANN model can be used to establish selection criteria and breed standards for Sujiang pigs.