APPLICATION OF THE CATBOOST ALGORITHM WITH OPTUNA OPTIMIZATION FOR CLASSIFYING THE ETIOLOGY OF ACUTE FEVER USING SMOTE-NC AS A DATA CLASS BALANCER

Ramadhan, Raihan (2026) APPLICATION OF THE CATBOOST ALGORITHM WITH OPTUNA OPTIMIZATION FOR CLASSIFYING THE ETIOLOGY OF ACUTE FEVER USING SMOTE-NC AS A DATA CLASS BALANCER. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) Cover.pdf Download (945kB)
	Text (Bab 1) Bab 1.pdf Download (206kB)
	Text (Bab 2) Bab 2.pdf Restricted to Repository staff only until 3 September 2028. Download (842kB)
	Text (Bab 3) Bab 3.pdf Restricted to Repository staff only until 3 September 2028. Download (1MB)
	Text (Bab 4) Bab 4.pdf Restricted to Repository staff only until 3 September 2028. Download (1MB)
	Text (Bab 5) Bab 5.pdf Download (194kB)
	Text (Daftar Pustaka) Dapus.pdf Download (215kB)
	Text (Lampiran) Lampiran.pdf Restricted to Repository staff only Download (1MB)

Abstract

This research addresses class imbalance and symptomatic overlap in the multiclass classification of acute fever etiology by integrating CatBoost and SMOTE-NC with hyperparameter optimization via Optuna, implemented as a mobile-based decision support system. The secondary dataset, derived from the AFIRE Study (2013–2016), yielded 801 clean samples across four target classes: Dengue, Rickettsia spp., Other bacterial, and Other viral. Using Stratified 5-Fold Cross Validation, the optimal CatBoost+SMOTE-NC+Optuna model achieved an average accuracy of 74.16%, Macro Recall of 64.74%, and Macro F1-Score of 64.08%, with the strongest performance on Dengue (F1=0.87) and more moderate results for the minority Rickettsia spp. class (F1=0.51). SMOTE-NC significantly improved Macro Recall without a statistically significant loss in Accuracy. Statistical testing confirmed CatBoost significantly outperformed XGBoost across all metrics (p<0.05), while showing comparable performance to Random Forest (p>0.05); CatBoost was ultimately selected for its architectural efficiency and inference time of just 0.024 ms. A Brier Score-based calibration analysis further confirmed the reliability of the predicted probabilities. The final model was deployed in a client-server architecture (FastAPI backend, Flutter frontend) and successfully passed black-box testing.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Via, Yisti Vita	NIDN0025048602	yistivia.if@upnjatim.ac.id
Thesis advisor	Rakhmadi, Ardhon	NIDN9990610207	ardhon.rakhmadi.fasilkom@upnjatim.ac.id

Subjects:

Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Q Science > QA Mathematics > QA76.6 Computer Programming

Divisions:

Faculty of Computer Science > Departemen of Informatics

Depositing User:

Raihan Ramadhan

Date Deposited:

07 Sep 2026 01:36

Last Modified:

07 Sep 2026 01:36

URI:

https://repository.upnjatim.ac.id/id/eprint/59938