Comparison of the Predictive Capabilities of Several Data Mining Algorithms and Multiple Linear Regression in the Prediction of Body Weight by Means of Body Measurements in the Indigenous Beetal Goat of Pakistan

Abstract

The main goal of this study was to establish the algorithm with the best predictive capability among classification and regression trees (CART), chi-square automatic interaction detector (CHAID), radial basis function (RBF) networks and multilayer perceptrons with one (MLP1) and two (MLP2) hidden layers in body weight (BW) prediction from selected body measurements in the indigenous Beetal goat of Pakistan Moreover, the results obtained with the data mining algorithms were compared with multiple linear regression (MR). A total of 205 BW records including one categorical (sex) and six contmuous (head girth above eyes, neck length, diagonal body length, belly sprung, shank circumference and rump height) predictors were utilized The Pearson correlation coefficient between the actual and predicted BW (r) and root-mean-square error (RMSE) were used as goodness-of-fit criteria, among others A 10-fold-cross validation was applied to tram and test CART, CHAID and ANN and to estimate MR coefficients. The most significant BW predictors were sex, rump height, shank circumference and head girth The r value ranged from 0.82 (MLPI) to 0 86 (RBF and MR) The lowest RMSE (3.94 kg) was found for RBF and the highest one (4.49 kg) for MLPI In general, the applied algorithms quite accurately predicted BW of Beetal goats, which may be helpful in making decisions upon standards, favourable drug doses and required feed amount for animals. The ascertainment of the body measurements associated with BW using data mining algorithms can be considered as an indirect selection criterion for future goat breeding studies.