Recently, as information technology continues to develop, data mining methods, which are techniques for obtaining useful and understandable information from large numbers of data, have become more important in the fields of medical and pharmaceutical sciences. Such methods are being adopted for studies involving large numbers of data, such as DNA microarray or epidemiological data. However, most data mining methods are unsuitable in these fields because validating attributes such as the risk factors of medications are not easy to validate using models. Since understanding the relationships between treatments and their effects is especially important in medical and pharmaceutical sciences, practitioners in these fields tend to avoid such unsuitable methods for data analyses, even when large volumes of data have to be analyzed. The decision tree is one of these methods. In this study, we propose a novel procedure which enables users to clarify the relationships between attributes and classification results by tree models using resampling methods. Our new procedure has a function to obtain information about the significance of the attributes for tree models. In addition, this method can extend the applicability of decision tree-like methods.
Read more
1