Open AccessAnalysis of Attribute Reduction Effectiveness on The Naive Bayes Classifier Method
Author(s) -
D. Syafira,
Saib Suwilo,
Poltak Sihombing
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1566/1/012060
Subject(s) - bayes error rate , naive bayes classifier , bayes' theorem , probabilistic classification , artificial intelligence , bayes classifier , computer science , principal component analysis , machine learning , pattern recognition (psychology) , data mining , classifier (uml) , bayesian probability , mathematics , support vector machine
Naïve Bayes is a prediction method that contains a simple probabilistic based on the application of the Bayes theorem (Bayes rule) with the assumption that the dependence is substantial. Classification is a technique in data mining to form a model from a predetermined data set. Data mining techniques are the choices that can overcome in solving this problem. Therefore, researchers make a comparison of Naïve Bayes by modifying using PCA. Improving the comparative performance of the Naive Bayes method by weighting using Principal Component Analysis (PCA) and Naive Bayes was carried out in this study. Research conducted on the classification model of Naïve Bayes (PCA + Naïve Bayes) using the Iris dataset that simplified into two attributes, four classes and 147 instances with an accuracy rate of 97.78% with a classification error rate of 2.22%. Meanwhile, the Conventional Naïve Bayes classification model uses four attributes with four classes from the Iris dataset, which has an accuracy rate of 95.45% with a classification error rate of 4.55%. Based on the test results of the classification model, it can be concluded that the success of the PCA can be used as a reference to improve the accuracy performance of the Naïve Bayes classification model.