Skip to content
SDGs Center UDINUS

Evidence record · UDINUS · SDGs

Enhancing Prediction Accuracy in an Imbalanced Dataset of Dengue Infection Cases Using a Two-layer Ensemble Outlier Detection and Feature Selection Technique

Rektorat UDINUS 2024 Published 2026-10-30

Evidence record · UDINUS · SDGs

Unit
Rektorat UDINUS
Period
2024
Goals
Record
547fd9c2-0c6a-4861-8ffa-88a2a43d46f2
Published
2026-10-30

Universitas Dian Nuswantoro (UDINUS), with Fahmi, Amiq; Purwitasari, Diana; Sumpeno, Surya; and Purnomo, Mauridhi Hery as participants, conducted research on improving prediction accuracy in an imbalanced real-world health dataset of dengue infection cases. The research addressed noise and outlier data in real-world datasets. In classification datasets, the presence of outliers can make predictive modelling more difficult and can result in inaccurate and lower predictive modelling performance. The activity therefore focused on detecting and removing outliers in a large and imbalanced healthcare dataset.

The proposed approach used a two-layer ensemble method called IsFLOF. IsFLOF combined an isolation forest, identified as IsF, and a local outlier factor, identified as LOF, to find and eliminate global and local outliers. The approach addressed the stated limitations of the individual methods. IsF is sensitive to global outliers but vulnerable to local outliers, while LOF performs well in local outlier detection but has high complexity. After outlier elimination produced a dataset with correctly measured value distributions, resampling was conducted to prevent prediction bias caused by imbalanced instance data in a multi-class setting. Insignificant features were then filtered to refine the dataset.

Eight machine learning algorithms were used to test robustness and effectiveness. AdaBoost combined with features selected through the Fast Correlation-Based Filter achieved 93.5% accuracy in training and 95.1% accuracy in testing. The proposed method was also tested and compared with recent methods using a public dataset of imbalanced hypothyroid cases, where it showed higher and more acceptable prediction accuracy than the original and synthetic data.

Reference https://www.scopus.com/pages/publications/85188175562?origin=resultslist