Pattern discovery by residual analysis and recursive partitioning

In this paper, a novel method of pattern discovery is proposed. It is based on the theoretical formulation of a contingency table of events. Using res idual analysis and recursive partitioning, statistically significant events are identified in a data set. These events constitute the important inform ation contained in the data set and are easily interpretable as simple rule s, contour plots, or parallel axes plots. In addition, an informative proba bilistic description of the data is automatically furnished by the discover y process. Following a theoretical formulation, experiments with real and s imulated data will demonstrate the ability to discover subtle patterns amid noise, the invariance to changes of scale, cluster detection, and discover y of multidimensional patterns. It is shown that the pattern discovery meth od offers the advantages of easy interpretation, rapid training, and tolera nce to noncentralized noise.