Comparison of Artificial Neural Network and Decision Tree Algorithms used for Predicting Live Weight at Post Weaning Period From Some Biometrical Characteristics in Harnai Sheep

The present investigation deals with comparing performance of different data mining algorithms viz., CHAID, EXHAUSTIVE CHAID, CART and ANN, in order to predict body weight from biometrical characteristics for a total of 161 Harnai lambs changing in age between 6-9 months. Minimum numbers of lambs in parent and child nodes were set 20:10 for first three algorithms in order to improve their predictive ability. Considered as an outcome variable, body weight were predicted from explanatory variables, linear body measurements; namely, body length (BL), withers height (WH), chest girth (CG), paunch girth (PG), face length (FL), length between ears (LBE), length of ears (EARL), width (FTW) and length (FTL) of tail, and sex factor. To find out the best one among aforenamed four algorithms, model quality criteria like coefficient of determination (R2%), adjusted coefficient of determination (Adj-R2%), coefficient of variation (CV%), SD ratio (the ratio of SD of errors to SD of the actual dependent variable values), Root Mean Square Error (RMSE), Relative Approximation Error (RAE) and the Pearson correlation between actual and predicted values were estimated. All the qualified algorithms were very sufficient; therefore, the use of Exhaustive CHAID in the prediction of body weight presented the best fit in both model quality criteria and more proper decision tree diagram visually, which may provide some advantageous in exhibition of some breed standards of the Harnai sheep