Open AccessEvaluation of feature selection using information gain and gain ratio on bank marketing classification using naïve bayes
Author(s) -
Budi Prasetiyo,
Alamsyah Alamsyah,
Much Aziz Muslim,
Niswah Baroroh
Publication year - 2021
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 - 10.1088/1742-6596/1918/4/042153
Subject(s) - information gain ratio , feature selection , information gain , ranking (information retrieval) , naive bayes classifier , entropy (arrow of time) , computer science , selection (genetic algorithm) , mutual information , feature (linguistics) , artificial intelligence , machine learning , pattern recognition (psychology) , data mining , support vector machine , linguistics , physics , philosophy , quantum mechanics
One of the efforts of banks to do marketing is by telephone to offer their products, such as deposits. There are many variables that influence whether the customer decides to subscribe or not. In this study, we present a comparison of feature selection from high features dataset. We use a bank marketing dataset which has 20 features and consists of 4,119 instances. We consider 2 ranking methods entropy-based, namely Information Gain (IG) and Gain Ratio (GR). In our experiment, we classified the various selected based on the ranking of the selected features using Naïve Bayes. We show that the selection of different features is important for classification accuracy. The different combinations of feature selection can affect the accuracy results.