Tendency Discovery from Incident Reports with Free Format Using Self Organizing Map

Abstract

 The purpose of this study is construction of knowledge discovery method from medical documents with free format. As the first step of this study, this paper discussed the keyword map generation method using Self Organizing Map(SOM) from incident report with free format to discover new trends or features. In particular, this paper proposed (1) the keyword extraction method using morphological analysis tools and heuristics rules from incident reports and (2) the coding method for SOM using the extracted keywords. In the proposed method, the some items in the incident reports (e.g. occurred time, category of occurrence, medical staff’s name etc.) were not employed to generate the keyword map but only free format sentences were did. By the proposed methods, this paper showed that appropriate keywords in the incident reports were extracted, and the effective keyword map was automatically generated. An example of generated keyword map showed some connections of keywords visually and gave the possibility for knowledge discovery(i.e. trends or features underlaid in reports and difficult to find from each document) using the proposed method. Moreover, this paper described the problems and the future works of the proposed method, and discussed the possibility of the proposed method to develop the risk management support system.