The aim of this investigation was to develop a prediction equation for fattening final live body weight
from several body measurements and fattening period of native, crossbred and exotic breeds. For this aim,
a total of 103 young bulls were used. In the prediction of fattening final live weight as an output variable,
several continuous predictors evaluated in the current study were: withers height (WH), back height (BH),
front rump height (FRH), back rump height (BRH), body length (BL), back rump width (BRW), chest
depth (CD) and chest circumference (CC). Also, the breed factor was considered as a nominal predictor
and fattening period (FP) was accepted as an ordinal predictor. To obtain the prediction equation, the
results of Multivariate Adaptive Regression Splines (MARS) data mining algorithm as a non-parametric
regression technique was implemented. To measure predictive accuracy of MARS, model evaluation
criteria such as coefficient of determination (R2), adjusted coefficient of determination (R2
ADJ), SDRATIO and
Pearson coefficient (r) between actual and predicted values in fattening final live weight were calculated.
To reveal the highest predictive ability in the MARS algorithm, numbers of terms and basis functions
were set at 21 and 45 where order of interactions was three. Except for CD, other predictors were entered
into MARS model. MARS showed very high predictive capability (R2=0.9717, R2
ADJ=0.9643, SDRATIO=
0.168 and r=0.986) for the data evaluated in the investigation. Also, GCV value of the MARS prediction
equation was found as 409.83. In conclusion, it could be suggested that a very reliable prediction equation
with the predictive accuracy of nearly 100 (%) was developed in practice by using MARS data mining
algorithm, which a quite remarkable tool in the prediction of fattening final live weight with interaction
effects of predictors and in description of breed standards, in the development of breeding strategies and
especially in the detection of ideal fattening period for each breed under the condition.