DFP-SEPSF: A dynamic frequent pattern tree to mine strong emerging patterns in streamwise features

Mining a minimal set of strongly predictive emerging patterns from a high dimensional dataset is a challenging issue for making an accurate emerging pattern classifier. The problem becomes even more severe when features are not available as a whole; in this scheme, features are emerged one by one instead of having all features at hand before learning process gets started. In this study, we propose a novel dynamic structure to construct the frequent pattern tree on arrival of new features and to mine emerging patterns online. DFP-SEPSF, a Dynamic Frequent Pattern tree to mine Strong Emerging Patterns in Streamwise Features, offers an efficient bottom up approach to construct an Unordered Dynamic Frequent Pattern tree (UDFP-tree) and an Ordered Dynamic Frequent Pattern tree (ODFP-tree). Moreover, the proposed framework extracts Strong Emerging Patterns (SEPs) to build an accurate and fast classifier that can deal with noise. Our experimental evaluations indicate the effectiveness of the proposed approach in comparison with other state-of-the-art methods, in terms of predictive accuracy, emerging pattern numbers, and running time.