Comparison of artificial neural network and multiple linear regression for prediction of first lactation milk yield using early body weights in Sahiwal cattle

Abstract

A comparative study of connectionist network (also known as artificial neural network, ANN)) and multiple regression is made to predict the first lactation 305 days or less milk yield (FL305DMY) from early body weights in Sahiwal cattle. A multilayer feed forward network with back propagation of error learning mechanism was used for prediction. Data collected from 221 Sahiwal heifers were partitioned into 2 data sets namely training data set comprising 75% data to build the neural network model and test data set comprising 25% to test the model. Early body weights, viz. birth weight, body weights at 6, 12, 18, 24, 30 months and weight at first calving were used as input variables and FL305DMY was considered as output variable. The same training and test data sets were used for multiple linear regression analysis (MLR). The prediction efficiency of both models was compared using the R2 value and root mean square error (RMSE). The accuracy of prediction from both the models was observed to be very low. However, the accuracy of prediction was comparatively higher from ANN model (6.87%) than MLR model (3.32%) for test set of data. The comparatively low value of RMSE and high value of R2 in case of connectionist network in comparison of MLR model shows that connectionist network model is a better alternative to the conventional regression model.