Comparative study of artificial neural network algorithms performance for prediction of FL305DMY in crossbred cattle

In the current investigation, records of 1092 crossbred cattle (Vrindavani) were collected at Indian Veterinary Research Institute, Izatnagar. Crossbred cattle's First Lactation 305-Day Milk Yield (FL305DMY) was predicted using three separate Artificial Neural Network (ANN) algorithms, and their performance was evaluated. Each algorithm's efficiency was measured and evaluated on the basis of the coefficient of determination and Root Mean Square Error (RMSE). Two different set of inputs were used in the analysis to predict the yield of milk. The first set of inputs comprised of a record of test day milk yields together with age at first calving (AFC) and peak yield (PY) and a second set comprised of monthly milk yield records, AFC and PY. Three ANN algorithms used for training were Levenberg-Marquardt (LM), Bayesian Regularization (BR) and Scaled Conjugate Gradient (SCG). Every algorithm was evaluated using four separate data sets for the training test (66.66:33.33, 75:25, 80:20, and 90:10). BR reached 79.89% best accuracy with 16.89% lowest RMSE value for first input set and 82.67% accuracy with 14.45% RMSE value for second input set-2. Therefore, BR algorithm can be used to predict FL305DMY in Crossbred cattle as it demonstrates higher accuracy over LM and SCG algorithm.