Knowledge Discovery in Mortality Records: An Info-Fuzzy Approach

In this chapter, we analyze a data set, which has been provided by the Israel Ministry of Health. It includes the records of all Israeli citizens who passed away in the year 1993. The death cause (medical diagnosis) of each person is defined by the international, 6-digit code (ICD-9-CM). We start the chapter with the description of the data pre-processing operations, such as transformation and cleaning of the original attributes. An information-theoretic data mining algorithm, developed in our previous work, is applied to the pre-processed dataset. The algorithm includes discretization, dimensionality reduction, and rule extraction. The original dataset is extended to a fuzzy relational table by adding a new, fuzzy attribute: the reliability degree of the recorded diagnosis. Our approach to the calculation of the reliability degree is based on the Computational Theory of Perception (CTP). Most records having the lowest reliability degree are shown to represent exceptional and possibly inaccurate information.