| Autor: |
Efendi, Achmad, Amrullah, Ahmad A. N., Fitriani, Rahma, Rahayudi, Bayu |
| Předmět: |
|
| Zdroj: |
International Journal of Agricultural & Statistical Sciences; 2021 Suppl, Vol. 17, p925-938, 14p, 7 Diagrams, 13 Charts |
| Abstrakt: |
The Multivariate Adaptive Regression Spline (MARS) model is one of the methods in nonparametric regression, which is popularly used for solving prediction and classification problems. The problem that arises in the classification method is the imbalance of the number of observations between classes on the response variable. The MARS classification model is generally effective in classifying this unbalanced data as it is an adaptive model that can form a model by adjusting existing data patterns. To reduce the misclassification of the minority class on unbalanced data and to increase the accuracy or classification strength of the MARS model, two methods can be also used, Bootstrap Aggregating (Bagging) and Synthetic Minority Over-sampling Technique (SMOTE). This study aims at determining the accuracy of the classification using the MARS, Bagging-MARS and SMOTE-MARS methods on simulation data that has an imbalance of the number of observations between classes with the proportion of minority classes of 10%, 15% and 20%. The results of this study indicate that the classification accuracy of the Bagging-MARS method has the best classification power compared to the MARS and SMOTEMARS methods by looking at the accuracy, sensitivity and APER values generated at each level of the proportion of minority classes. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
