A Novel Fuzzy Linguistic Fusion Approach to Naive Bayes Classifier for Decision Making Applications

Abstract

Naive Bayes is one of the most widely used classifier algorithms in various data mining problems. The performance of the Naïve Bayes Classifier is comparable to other classifiers as it yields impressive results in multiple applications. An increase in the performance of the Naive Bayes Classifier is possible by identifying and forming segments of the data handled by the classifier. In this paper, a novel fuzzy-based fusion approach to selected quantitative features is proposed. The approach is used to improve the prediction accuracy of the Naive Bayes Classifier (NBC). The linguistic computing model with fusion operators, using ranked indexes of the linguistic terms in the dataset is made use in this proposed approach. Fuzzy values are generated only for the numerical attributes in the initial phase using 2-tuple linguistic computations. The equivalent real value computations are performed in order to express the results in the initial domain of the expression. These computations ensure improved comprehensiveness of the results of the classifier. The model incorporates the concepts of linguistic terms, fuzzy logic, fusion methods, and aggregation operations to the classical Naïve Bayes Classifier. Such incorporation is used to improve the performance of the classifier in various decision-making applications. The proposed model is validated using a standard benchmark dataset–Stat log Heart disease dataset. It is obtained from the UCI Machine Learning Repository. The proposed Linguistic Fuzzy Naive Bayes Classifier showed better accuracy compared to the Simple Naive Bayes Classifier performance.