A brief introduction to the field of data mining

Abstract

The authors were involved in a study to determine a pattern in drug usage by the firefighters in Los Angeles. Using pharmacy records, both doctors' prescriptions and over the counter drugs, the most typical types of drugs, the amount used and the length of the drug's usage were detected by the means of basic statistical and clustering methods. The main purpose was to improve the quality of the health care institution, while minimizing costs of drugs. Thus, the problem was not only finding a pattern but also the optimization of a quality function. The basic methods in data mining came from statistics (regression, analysis of variance, cluster or classification analysis, and time series), operations research (decision trees, genetic algorithms and neural networks), and/or algorithms in computer science. Three basic approaches in data mining are presented in the paper: classification/clustering; neural networks; and decision trees.