Comparative efficiency of artificial neural networks and multiple linear regression analysis for prediction of first lactation 305-day milk yield in Sahiwal cattle

Abstract

An investigation was carried out on 12,854 fortnightly test day milk yields records of first lactation pertaining to 643 Sahiwal cows sired by 51 bulls spread over 49 years located at the National Dairy Research Institute, Karnal. The comparison was made between the relative efficiency of multiple linear regression analysis and artificial neural network (ANN) for prediction of first lactation 305d milk yield (FL305DMY) in Sahiwal cows. Artificial Neural Network was trained using three back propagation algorithms viz. Bayesian regularization (BR), Scaled conjugate gradient (SCG) and Levenberg–Marquardt (LM). Further, these three algorithms were compared using four sets of training and test data sets at 66.67–33.33%, 75–25%, 80–20% and 90–10%. It has been found that the coefficient of determination of the models was increased with the addition of test day milk yields as input variables. It was inferred from the study that artificial neural network was better than the multiple linear regression analysis to predict FL305DMY with more than 80% accuracy by almost all the models at an early stage i.e. by 111th day of the lactation having lesser value of RMSE than MLR. Therefore, it is recommended that ANN can be a potential tool for the prediction of the first lactation 305-day milk yield in Sahiwal cows.