Skip to content
SDGs Center UDINUS

Evidence record · UDINUS · SDGs

Evaluation of Feature Selection Using Wrapper for Numeric Dataset with Random Forest Algorithm

Rektorat UDINUS 2021 Published 2026-10-30

Evidence record · UDINUS · SDGs

Unit
Rektorat UDINUS
Period
2021
Goals
Record
f1202324-e9df-47ab-9a48-30e9ea622a64
Published
2026-10-30

Universitas Dian Nuswantoro conducted research on feature selection using a wrapper method for a numeric dataset used for early diabetes prediction. The activity involved UDINUS researchers Nugroho, Arie; Fanani, Ahmad Zainul; and Shidik, Guruh Fajar. The research addressed preprocessing in machine learning, particularly dimensionality reduction through feature selection. The stated purpose was to identify relevant features and remove features that were not relevant from a dataset that had previously been transformed into a numeric dataset.

The activity applied forward and backward selection within the wrapper method. These selection approaches were combined with a random forest algorithm and cross validation. Random forest was described as an enhancement of decision trees, consisting of a group of trees that may produce the same or different results, with the most results used as the final result. The research assessed the results of feature selection against processing without feature selection for the numeric dataset.

The reported findings indicate that wrapper-based feature selection produced higher accuracy than processing without feature selection and reduced the number of features. Sequential forward selection achieved 98.84% accuracy using 11 selected features. Sequential backward selection achieved 99.03% accuracy using the same number of selected features. The reduced feature set was also reported to reduce tree complexity and the time required in the mining process.

The activity has a health-related research context because it used an early diabetes prediction dataset. It also addressed feature relevance, predictive processing, and dimensionality reduction through the application of wrapper selection, random forest, and cross validation. This frames the activity as institutional research by UDINUS concerning methods for processing a numeric dataset associated with early diabetes prediction.

Reference https://www.scopus.com/pages/publications/85118929554?origin=resultslist