An Improved Cellular Automata-Based Classifier with Soft Decision

Abstract

Classification has been successfully applying in problems in a variety of fields, such as science, business, engineering, and industry. Unfortunately, the classifier coping with nonconforming binary patterns are rare. To deal with nonconforming pattern in binary Cellular Automata-based Classifier (CAC) had been proposed. However, CAC faces several limitations that need to improve. First, the rule ordering process in CAC which used Genetic Algorithm (GA) is unable to handle high dimensional complex problems. Second, finding decision boundaries is quite rough when dealing with ambiguous data. To deal with these problems, therefore, we propose a new classifier, called Cellular Automata-Based Classifier with Soft Decision (CAS). We replace the GA with the promising optimization algorithm, called Butterfly Optimization, for the rule ordering process. Subsequently, we improve the classification performance by augmenting a Soft-Decision step. This Soft-Decision step uses the pruning method to create a soft decision table, which efficiently serves for filtering useless data. Finally, to verify the classification performance of the proposed method, ten datasets consisting of conforming and nonconforming patterns are experimented in comparison with the promising classifiers including CAC, Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Naïve Bayes, and Deep Learning using K-fold cross-validation. In this regard, CAS provides the promising results.