Vol. 12, Special Issue 12 (2023)

Author(s):
K Andrew Jabakumar, JK Chaudhary, N Shyamsana Singh, TC Tolenkhomba, Girin Kalita and Ranjana Goswami

Abstract:
The study was conducted to determine the most adequate regression method among Multiple Linear Regression (MLR), Ridge Regression (RR), Least Absolute Shrinkage and Selection Operator (LASSO), Elastic Net (EN), CART (Classification and Regression Tree) and Random Forest Regression (RF) for the prediction of egg weight (EW) from various internal and external egg quality characteristics namely, Shape index (SI), Yolk height (YH), Yolk index (YI), Albumen height (AH), Haugh unit (HU), Albumen index (AI), Yolk ratio (YR), Albumen ratio (AR), Shell weight (SW) and Shell thickness (ST) in quail eggs. In this experiment, a total of 100 eggs were collected. Various goodness of fit criteria namely, coefficient of determination (R²), Adjusted coefficient of determination (Adj. R²), Mean Squared Error (MSE), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Pearson correlation coefficient were estimated for describing the most adequate regression method in terms of the predictive performance. In the study, SW was identified to be the most important predictor of egg weight in Japanese quails. The highest predicted EW (10.24 g) was obtained from eggs with SW≥1.21 g. The phenotypic correlation between egg weight and shell weight was significantly positive and high (0.55) in Japanese quails. On comparing the predictive performance of the six regression methods employed in the current study using different model evaluation criteria, the highest Adj. R² value was obtained for the MLR (97.66%) which emerged to be the best suited model due to the existent of low multicollinearity among the predictors in the Japanese quail dataset. The shell weight had high positive correlation (0.55) with egg weight.