A Study on Regression Tree for Count Data with a Data mining Application on Industrial Accidents

펼치기

Decision tree as one of many data mining techniques is a popular approach for segmentation, classification and prediction by applying a series of simple rules. In general, to analyze continuous target variable, we use F-statistics or variance reduction criterion to find the best split. But these methods are only appropriate to a continuous target variable. If the target variable is discrete, especially count data, above criteria couldn’t give a good result to analyst because of its attribute. In this paper, we will propose a decision tree for count data, rare event, using maximum poisson likelihood as split criterion and using Korean industrial accident data sets, we will compare the performance of the split criteria.