Classification Using Jumping Emerging Patterns and Cosine Similarity

Classification is a common task in Machine Learning and Data Mining. Jumping Emerging Patterns have been applied for classification in different contexts with good results and the advantage of to be easily understandable for users. In this work we propose the use of cosine similarity measure to select the patterns which will be used to predict the classes in the classification process. Two versions of the algorithm were proposed and tested with four different parameter values in 21 datasets. The results were compared with three frequently used classification algorithms from the literature and proposed algorithms showed a promising results achieving in some the datasets best results than C4.5