View article

In recent years, researchers have proposed many techniques for detecting outliers in data sets. However, most of these techniques assume that the data set is static and consists of homogeneous attribute types. However, these assumptions do not hold for many real-world data sets. To address these weaknesses, we present a technique for outlier detection in dynamic mixed-attribute data. Our technique is capable of finding outliers in a single pass of the data, and can do so with low memory requirements. Our approach uses a combination of classifiers and statistical tests to discover anomalous values of categorical and continuous attributes. Our empirical results demonstrate that while our technique only shows marginal improvements in detection rates, its execution speed and memory usage are far better than those of current state-of-the-art outlier detection techniques.