Supervised Evaluation of Top-k Itemset Mining Algorithms

A major mining task for binary matrixes is the extraction of approximate top-k patterns that are able to concisely describe the input data. The top-k pattern discovery problem is commonly stated as an optimization one, where the goal is to minimize a given cost function, e.g., the accuracy of the data description. In this work, we review several greedy state-of-the-art algorithms, namely Asso, Hyper+, and PaNDa(+), and propose a methodology to compare the patterns extracted. In evaluating the set of mined patterns, we aim at overcoming the usual assessment methodology, which only measures the given cost function to minimize. Thus, we evaluate how good are the models/patterns extracted in unveiling supervised knowledge on the data. To this end, we test algorithms and diverse cost functions on several datasets from the UCI repository. As contribution, we show that PaNDa(+) performs best in the majority of the cases, since the classifiers built over the mined patterns used as dataset features are the most accurate.