Lameness prediction in Karan Fries cross-bred cows using decision tree models

Decision tree algorithms were used to develop predictive models for lameness based on percent of body weight distribution to individual legs of Karan Fries cross bred cows. To develop prediction model, 589 Karan Fries cows data were recorded for the health status (lame or healthy) as target variable and other non-genetic variables such as percent of body weight distributed to individual legs (using load cell platform), parity (1 to 10), status of pregnancy ( 181days and nonpregnant), status of lactation ( 121days and dry), and daily milk yield were used as input variable. The predictive models were optimized based on classification of health status as well as average square error (ASE) to predict lameness in cows. The performance of both the models were evaluated based on accuracy rate, sensitivity, specificity, misclassification rate and average square error under different data partition schemes(60∶40, 70∶30 and 80∶20). This study reveals that decision tree model optimized for average square error (ASE) with 80∶20 data partition scheme had minimum average square error (ASE) (0.146) and maximum sensitivity (79.67%) to predict the lame cow as lame.