Milk yield prediction in Pakistani camels using selected data mining algorithms

The aim of the present study was to predict lactational milk yield in eight indigenous camel breeds of Pakistan. Selected productive (hair production and adult body weight) and reproductive (age at puberty, age at first breeding, gestation length, dry period, and calving interval) traits were used to develop selected data mining predictive models. The models included classification and regression trees (CART), chi-square automatic interaction detection trees (CHAID), exhaustive CHAID, multivariate adaptive regression splines (MARS), multilayer perceptrons (MLP), and radial basis function networks (RBF). The highest Pearson correlation coefficients between the observed and predicted values were obtained with CART, CHAID, EXCHAID, MARS and MLP. MARS was also characterised by the lowest root mean square error, standard deviation ratio, mean absolute percentage error, and mean absolute deviation. The most important identified predictors of lactational milk yield were camel breed, adult body weight, age at first breeding, calving interval and dry period (depending on the model). The applied data mining methods allowed for the accurate prediction of lactational milk yield in camels.