Text Mining for Nursing Care Service Improvement
Tomohiro Minezaki, Sozo Inoue,
International Symposium on Applied Engineering and Sciences (SAES2016)
This paper describes the result of mining accident reports in nursing facilities for the purpose of obtaining the accident-prone situation. In addition, we discuss about measures for accident reduction on the basis of the results. In order to detect the pattern of the accident at the nursing facility, we combined text mining and data analysis of the accident reports retrospectively. First, we took the accident detail part from accidents report, split words using the MeCab, and made a word-document matrix. Next, we performed k-means clustering method to the 2701 words, and classified into 50 clusters. Then, we created a multi-variate data, combining the words-document matrix and the attributes of the accident reports. Moreover, we estimated the attribute which we set as an objective variable using the Random Forest method with this multivariate data. Finally, we made a decision tree using the explanatory variables of high importance word clusters and attributes obtained by the Random Forest method, and analyzed the details of the attributes. As a result, it was found the nursing accident trends, such as the type of relationship between the accident-prone situations and the purpose of the action.