Making Use of the Most Expressive Jumping Emerging Patterns for Classification

Classification aims to discover a model from training data that can be used to predict the class of test instances. In this paper, we propose the use of jumping emerging patterns (JEPs) as the basis for a new classifier called the JEP-Classifier. Each JEP can capture some crucial difference between a pair of datasets. Then, aggregating all JEPs of large supports can produce a more potent classification power. Procedurally, the JEP-Classifier learns the pair-wise features (sets of JEPs) contained in the training data, and uses the collective impacts contributed by the most expressive pair-wise features to determine the class labels of the test data. Using only the most expressive JEPs in the JEP-Classifier strengthens its resistance to noise in the training data, and reduces its complexity (as there are usually a very large number of JEPs). We use two algorithms for constructing the JEP-Classifier which are both scalable and efficient. These algorithms make use of the border representation to efficiently store and manipulate JEPs. We also present experimental results which show that the JEP-Classifier achieves much higher testing accuracies than the association-based classifier of (Liu et al, 1998), which was reported to outperform C4.5 in general.