A Combined Learning Based Classifier Using Decision Tree and Logistic Regression

Overweight and obesity are pointed out as a serious threat to the health of people. According to a 2016 report by WHO, 1.9 billion people worldwide are overweight and 650 million are obese. Because the obesity causes various cardiovascular diseases, diabetes, hypertension, and musculoskeletal disorders, it is important to assess the risk of obesity in order to prepare its prevention policies. To deal with this problem, logistic regressions and decision trees are widely applied and then employed as primary classifiers. This paper intends to present a combined learning approach to classifying the risk of overweight and obesity. In our framework, logistic regression models are cascaded onto decision tree classifiers. This feature would be applicable to improving the predictive capability of primary classifiers. Experimental results, using an open dataset provided by Korea Disease Control and Prevention Agency, show that the proposed scheme improves the classification accuracy by 5% to 27%.