Application of Classification Tree Method to Determine Factors Affecting Somatic Cell Count in Holstein Cows

In the current study, the effect of phenotypic factors on Somatic Cell Count (SCC) was evaluated by using classification tree technique. The current study used a total of 1,972,031 test day records of SCC in parity 1 to 4 collected from 1,281 Iranian Holstein- Friesian cows’ herds through 2004–2013. The SCC records were converted to binary trait, defined as 1: If SCC≤ 200,000, and 2: Otherwise. The CART (Classification an‎d Regression Tree) algorithm for classification trees, with GINI index and Entropy function as the division criteria, was used to develop the tree. Statistical analysis was performed using ‘rpart’ package in R software. The constructed tree had 12 leaves and it was 6 levels deep. The results of classification tree procedure for ranking of importance of the variables responsible for the variation in SCC were, respectively, parity, test-day milk production, year of calving, season of calving, and days in milk (stage of lactation). Based on the obtained classification tree, different combination of variables associated with SCC could be identified. According to the classification tree, the lowest amount of SCC was expected in the group of cows that were in the 1st or 2nd parity; their test-day milk production was > 30 kg; they were calved from year 2004 to 2013; and their calving season was autumn or winter.