Detection of Difficult Calvings in Dairy Cows Using Boosted Classification Trees

The aim of this study was to apply boosted classification trees to dystocia detection in dairy cattle and to indicate its most important predictors. A total of 1742 calving records were used. Proportions of correctly detected difficult and easy calvings in a test set were 75.0%, 92.0% and 75.0%, 77.3%, for heifers and multiparous cows, respectively. The key predictors of calving difficulty were: pregnancy length, body condition score index, calving age, proportion of Holstein-Friesian genes (for heifers) as well as calving and calving-to-conception intervals, mean daily milk yield, fat and protein content in milk and 4%-fat corrected milk yield (for cows). The data mining method used in this study allowed us to obtain the models of good quality, confirmed by their relatively high effectiveness of detecting dystotic animals.