Interactive Exploration of Larger Pattern Collections: A Case Study on a Cocktail Dataset

We present a general method for employing interactive embedding techniques to enable an analyst to explore a larger collection of local patterns. The common idea among pattern-mining methods is to list descriptions of subsets of a dataset according to some interestingness measure. Because the space of all patterns in a dataset is exponentially large in the number of attributes, most pattern-mining algorithms reduce the output for the analyst to a small set of highly interesting and diverse patterns. However, by discarding most of the patterns, these methods have to make a trade-off between ruling out potentially insightful patterns and possibly drowning the analyst in results. We propose an alternative. To counteract information overload, we mine a rather large set of patterns and study this collection using an interactive embedding technique. Using this interactive, visually driven exploration technique, the analyst can develop an understanding of the patterns, their distribution, the concepts underlying them, and how they interrelate.