Discovery of temporal knowledge in medical time-series databases using moving average, multiscale matching, and rule induction

Abstract

Since hospital information systems have been introduced in large hospitals, a large amount of data, including laboratory examinations, have been stored as temporal databases. The characteristics of these temporal databases are: (1) Each record are inhomogeneous with respect to time-series, including short-term effects and long-term effects. (2) Each record has more than 1000 attributes when a patient is followed for more than one year. (3) When a patient is admitted for a long time, a large amount of data is stored in a very short term. Even medical experts cannot deal with these large databases, the interest in mining some useful information from the data are growing. In this paper, we introduce a combination of extended moving average method, multiscale matching and rule induction method to discover new knowledge in medical temporal databases. This method was applied to a medical dataset, the results of which show that interesting knowledge is discovered from each database.