Comparative Efficiency of Conventional and Connectionist Models in Prediction of First Lactation 305-day Milk Yield in Murrah Buffaloes

Background: Breeding programme is mainly structured and evaluated on the basis of first lactation 305-day milk yield (FL305DMY) of dairy animals. Early prediction of 305-day milk yield using test-day records is crucial for early evaluation of elite animals and to reduce the cost of data recording and animal rearing. Methods: In the study, the prediction efficiency of conventional models viz. centering date method (CDM), test interval method (TIM), ratio method (RM) and multiple linear regression (MLR) was compared with the newly evolved machine learning connectionist model named Artificial Neural Network (ANN). Data on 3,850 monthly test-day milk yield (MTDY) records of 809 Murrah buffaloes were utilized for the prediction of FL305DMY. The prediction efficiency of the models was compared based on absolute error, average error, root mean square error (RMSE) and their respective percentages. An attempt was thereafter made for the early-stage prediction of FL305DMY. Result: MLR was identified as the most accurate model with least error in prediction (4.19% RMSE), followed by ANN model (4.28% RMSE). The prediction accuracy for the regression equation incorporating all the 11 MTDY records was found to be 95.68 per cent. The optimal regression equation for early-stage prediction of FL305DMY consisted of four variables viz. MTDY-3 (65th day), MTDY-4 (95th day), MTDY-5 (125th day) and MTDY-7 (185th day) showed a R2 value of 87.02 per cent. It was inferred from the study that the most effective early-stage prediction of FL305DMY could be achieved by 185th day of lactation, offering a valuable tool for early and efficient selection of elite animals using monthly test-day milk yield records.