Efficient instance selection algorithm for classification based on fuzzy frequent patterns
A. Sabri Alvar, Mohammad Saniee Abadeh · 2016
Most of the instance selection methods seek to obtain subset of data for instance-based learning algorithms. These methods can improve classification performance, reduce memory requirements, and reduce execution time for these learning algorithms. In this paper, we introduce an instance selection algorithm (FF-IS) which is based on fuzzy frequent patterns and two thresholds. This method preserves appropriate border instances. The aim of this algorithm is to reserve important instances that are closer to border of the classes. We have used K-Nearest Neighbor (KNN) classifier to evaluate the performance of the proposed instance selection algorithm. We have compared our method with several well-known instance selection algorithms. Results indicate that this algorithm selects fewer and hence more representative instances. In comparison to other instance selection techniques, using the proposed instance selection algorithm enables the final KNN classifier to achieve better classification accuracy.