COMPARISON OF SMOTE RANDOM FOREST AND SMOTE K-NEAREST NEIGHBORS CLASSIFICATION ANALYSIS ON IMBALANCED DATA
In machine learning study, classification analysis aims to minimize misclassification and also maximize the results of prediction accuracy. The main characteristic of this classification problem is that there is one class that significantly exceeds the number of samples of other classes. SMOTE minor...
Main Authors: | , |
---|---|
Format: | Article |
Language: | English |
Published: |
Universitas Diponegoro
2023-04-01
|
Series: | Media Statistika |
Subjects: | |
Online Access: | https://ejournal.undip.ac.id/index.php/media_statistika/article/view/42755 |
_version_ | 1797394873991561216 |
---|---|
author | Jus Prasetya Abdurakhman Abdurakhman |
author_facet | Jus Prasetya Abdurakhman Abdurakhman |
author_sort | Jus Prasetya |
collection | DOAJ |
description | In machine learning study, classification analysis aims to minimize misclassification and also maximize the results of prediction accuracy. The main characteristic of this classification problem is that there is one class that significantly exceeds the number of samples of other classes. SMOTE minority class data is studied and extrapolated so that it can produce new synthetic samples. Random forest is a classification method consisting of a combination of mutually independent classification trees. K-Nearest Neighbors which is a classification method that labels the new sample based on the nearest neighbors of the new sample. SMOTE generates synthesis data in the minority class, namely class 1 (cervical cancer) to 585 observation respondents (samples) so that the total observation respondents are 1208 samples. SMOTE random forest resulted an accuracy of 96.28%, sensitivity 99.17%, specificity 93.44%, precision 93.70%, and AUC 96.30%. SMOTE K-Nearest Neighborss resulted an accuracy of 87.60%, sensitivity 77.50%, specificity 97.54%, precision 96.88%, and AUC 82.27%. SMOTE random forest produces a perfect classification model, SMOTE K-Nearest neighbors classification produces a good classification model, while the random forest and K-Nearest neighbors classification on imbalanced data results a failed classification model. |
first_indexed | 2024-03-09T00:26:05Z |
format | Article |
id | doaj.art-2def4a093d9a4b7abe4f9e8f63cfefb5 |
institution | Directory Open Access Journal |
issn | 1979-3693 2477-0647 |
language | English |
last_indexed | 2024-03-09T00:26:05Z |
publishDate | 2023-04-01 |
publisher | Universitas Diponegoro |
record_format | Article |
series | Media Statistika |
spelling | doaj.art-2def4a093d9a4b7abe4f9e8f63cfefb52023-12-12T02:27:52ZengUniversitas DiponegoroMedia Statistika1979-36932477-06472023-04-0115219820810.14710/medstat.15.2.198-20821856COMPARISON OF SMOTE RANDOM FOREST AND SMOTE K-NEAREST NEIGHBORS CLASSIFICATION ANALYSIS ON IMBALANCED DATAJus Prasetya0Abdurakhman Abdurakhman1PROGRAM STUDI MAGISTER MATEMATIKA, Departemen Matematika, Fakultas Matematika dan Ilmu Pengetahuan Alam, Universitas Gadjah Mada, Sekip Utara BLS 21 Yogyakarta 55281, IndonesiaDepartment of Mathematics, Gadjah Mada University, Indonesia, IndonesiaIn machine learning study, classification analysis aims to minimize misclassification and also maximize the results of prediction accuracy. The main characteristic of this classification problem is that there is one class that significantly exceeds the number of samples of other classes. SMOTE minority class data is studied and extrapolated so that it can produce new synthetic samples. Random forest is a classification method consisting of a combination of mutually independent classification trees. K-Nearest Neighbors which is a classification method that labels the new sample based on the nearest neighbors of the new sample. SMOTE generates synthesis data in the minority class, namely class 1 (cervical cancer) to 585 observation respondents (samples) so that the total observation respondents are 1208 samples. SMOTE random forest resulted an accuracy of 96.28%, sensitivity 99.17%, specificity 93.44%, precision 93.70%, and AUC 96.30%. SMOTE K-Nearest Neighborss resulted an accuracy of 87.60%, sensitivity 77.50%, specificity 97.54%, precision 96.88%, and AUC 82.27%. SMOTE random forest produces a perfect classification model, SMOTE K-Nearest neighbors classification produces a good classification model, while the random forest and K-Nearest neighbors classification on imbalanced data results a failed classification model.https://ejournal.undip.ac.id/index.php/media_statistika/article/view/42755machine learningclassificationsmoterandom forestk-nearest neighbors |
spellingShingle | Jus Prasetya Abdurakhman Abdurakhman COMPARISON OF SMOTE RANDOM FOREST AND SMOTE K-NEAREST NEIGHBORS CLASSIFICATION ANALYSIS ON IMBALANCED DATA Media Statistika machine learning classification smote random forest k-nearest neighbors |
title | COMPARISON OF SMOTE RANDOM FOREST AND SMOTE K-NEAREST NEIGHBORS CLASSIFICATION ANALYSIS ON IMBALANCED DATA |
title_full | COMPARISON OF SMOTE RANDOM FOREST AND SMOTE K-NEAREST NEIGHBORS CLASSIFICATION ANALYSIS ON IMBALANCED DATA |
title_fullStr | COMPARISON OF SMOTE RANDOM FOREST AND SMOTE K-NEAREST NEIGHBORS CLASSIFICATION ANALYSIS ON IMBALANCED DATA |
title_full_unstemmed | COMPARISON OF SMOTE RANDOM FOREST AND SMOTE K-NEAREST NEIGHBORS CLASSIFICATION ANALYSIS ON IMBALANCED DATA |
title_short | COMPARISON OF SMOTE RANDOM FOREST AND SMOTE K-NEAREST NEIGHBORS CLASSIFICATION ANALYSIS ON IMBALANCED DATA |
title_sort | comparison of smote random forest and smote k nearest neighbors classification analysis on imbalanced data |
topic | machine learning classification smote random forest k-nearest neighbors |
url | https://ejournal.undip.ac.id/index.php/media_statistika/article/view/42755 |
work_keys_str_mv | AT jusprasetya comparisonofsmoterandomforestandsmoteknearestneighborsclassificationanalysisonimbalanceddata AT abdurakhmanabdurakhman comparisonofsmoterandomforestandsmoteknearestneighborsclassificationanalysisonimbalanceddata |