Article Per Year (5 Year)
p-Index From 2020 - 2025
0.23
P-Index
This Author published in this journals
All Journal Prosiding Seminar Nasional Sisfotek (Sistem Informasi dan Teknologi Informasi)
Silvi Agustanti Bambang
Universitas AMIKOM Yogyakarta
Computer Science & IT

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

 1 
Kombinasi Algoritma Sampling dengan Algoritma Klasifikasi untuk Meningkatkan Performa Klasifikasi Dataset Imbalance Gagah Gumelar; Norlaila2; Quratul Ain; Riza Marsuciati; Silvi Agustanti Bambang; Andi Sunyoto; M. Syukri Mustafa
Prosiding SISFOTEK Vol 5 No 1 (2021): SISFOTEK V 2021
Publisher : Ikatan Ahli Informatika Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (368.284 KB)

Abstract

A class to be imbalanced when there is a class that has more data than other classes. A comparison between minority classes and the majority class is called Imbalance Ratio (IR). The greater the difference between the minority class and the majority class the value of the Imbalance Ratio (IR) is getting larger. Dataset imbalance in data mining is a serious problem. The application of the classification algorithm regardless of class balance resulted in a good prediction for the majority class and a neglected minority class. Therefore, in this research, the SMOTE algorithm was applied to balance the dataset. The study used 4 datasets with different Imbalance Ratio and used classification algorithms, C45, Naïve Bayes, K-NN, and SVM. Then compared before and after using SMOTE. The research results that have been done accuracy value and value G-mean Naïve Bayes algorithm is consistent with its performance at each level of imbalance ratio, before the implementation has no good performance, whereas after the implemented SMOTE algorithm Naïve Bayes has a consistent increase in accuracy. So it can be concluded that the combination SMOTE + Naïve Bayes most effectively used in the imbalance dataset with different levels in the scheme of 10 fold cross validation and 80% data testing tested as much as 50 times.
Co-Authors Andi Sunyoto Gagah Gumelar M. Syukri Mustafa Norlaila2 Quratul Ain Riza Marsuciati