APPLICATION OF DATA MINING ALGORITHMS FOR ESTIMATING LIVE BODY WEIGHT FROM LINEAR BODY MEASUREMENTS OF UKRAINIAN BEEF CATTLE BREED

The study was conducted to estimate live body weight of Ukrainian Beef Cattle (UBC) breed using Multivariate adaptive regression splines (MARS), and Classification and regression tree (CART) data mining algorithms.A total of 50 heifers at 12 months of age were used to collect live body weight (BW), Withers height (WH), rump height (RH), elbow join height (EJH), heart girth (HG), chest width (CW), width at the hip joints (WHJ) and width at the knee cups (WKC).The correlation results indicated that BW showed a highly positive correlation (P <0.01) with CW, EJH, and RH.MARS model showed that HG and CW were the most important traits for estimation of body weight of 12 months old UBC breed while CART model indicated WH, CW and RH were the most important traits.MARS and CART models were compared using Pearsons' correlation coefficient (r), coefficient of determination (R 2 ), adjusted R 2 (Adj.R 2 ), root-mean-square error (RMSE), Akaike information criteria (AIC), and coefficient of variance (CV).The results showed that CART was the best model with the highest R 2 (0.72), Adj.R 2 (0.72), and r (0.84) with the lowest RMSE (20.74),AIC (254.69), and CV (6.15).This study suggests that CART data mining algorithm might be used to determine breed standards of Ukrainian beef cattle breed for breeding program.