A Data Mining Method for Finding Hidden Relationship in Blood and Urine Examination Items for Health Check

Abstract

Our periodic health examination often describes whether each examination item in blood and urine takes in the reference range of each examination item and a simple summary report on checks in everyday life and the possibility of suspicious diseases. However, it uses n variable items such as AST(GOT), ALT(GPT) which are less correlated, and often includes expensive tumor markers. Therefore, this paper proposes a data mining method for finding hidden relationships between these items in order to reduce the examination fee and giving a report depending on individuals. Since low correlation coefficients are shown in most pairs of items over all clients, a set of item’s values in consecutive health examinations of each client is investigated for data mining. Four groups are formed according to the frequency taking outside the reference range in an item for three consecutive examinations, and average values of the other items included in each group are calculated in all pairs of items. The experiment results for three consecutive health examinations show that a lot of item pairs have positive or negative correlations between different frequencies with an item and the averages with the other item despite the fact that their correlation coefficients are small. The result shows both possible reducting of reducing the examination fee as inexpensive as possible and the possibility of a health-care report reflecting individuals.KeywordsBlood GlucoseReference RangeHealth ExaminationData Mining MethodRelation ExtractionThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.