Implementation of PSO-Random Forest in Classification of Heart Disease with Boruta Feature Selection and Data Balancing Technique SMOTE-Tomek

Habibi, Faisal Wildan (2026) Implementation of PSO-Random Forest in Classification of Heart Disease with Boruta Feature Selection and Data Balancing Technique SMOTE-Tomek. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text 21081010216-cover.pdf Download (1MB)
	Text (Bab 1) 21081010216-bab1.pdf Download (332kB)
	Text (Bab 2) 21081010216-bab2.pdf Restricted to Repository staff only until 22 July 2029. Download (751kB)
	Text (Bab 3) 21081010216-bab3.pdf Restricted to Repository staff only until 22 July 2029. Download (924kB)
	Text (Bab 4) 21081010216-bab4.pdf Restricted to Repository staff only until 22 July 2029. Download (3MB)
	Text (Bab 5) 21081010216-bab5.pdf Download (322kB)
	Text (Daftar Pustaka) 21081010216-daftarpustaka.pdf Download (326kB)
	Text (Lampiran) 21081010216-lampiran.pdf Restricted to Repository staff only Download (327kB)

Abstract

Heart disease is the major cause of death globally with 17 million annual cases, demanding accurate early detection. This study proposes a PSO-Random Forest approach with the application of Boruta feature selection and SMOTE�Tomek balancing on the UCI Heart Disease dataset (920 samples, 16 features, multi-class 0-4). The methodology includes preprocessing consisting of median/mode imputation, IQR capping, and encoding, PSO optimization with n_estimators parameters of 400-500 and max_depth of 10-20, and 4 scenarios: baseline, SMOTE-Tomek, Boruta, hybrid. The evaluation results show that the application of Boruta+SMOTE-Tomek achieves an optimal F1-score of 0.5917 (accuracy 0.6087, precision 0.6145), outperforming the baseline of 0.5894. Boruta selected 8 key features (age, trestbps, chol, thalach, dominant oldpeak), SMOTE-Tomek balanced the minority class (0:55%→50%). The 80:20 binary classification ratio scenario achieved 83.15% accuracy, and the multi-class scenario achieved 57.61%. PSO-RF proved robust overfitting compared to standard RF, with Boruta reducing noise by 48% and SMOTE-Tomek increasing minority recall by 12%. This approach contributes practically to ML�based cardiac diagnostic systems supporting early detection of disease stages in medicine. Keywords: PSO-Random Forest, Boruta, SMOTE-Tomek, Multi-Class Classification, UCI Heart Disease

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Wahanani, Henni Endah	NIP197809222021212005	henniendah@upnjatim.ac.id
Thesis advisor	Junaidi, Achmad	NIP197811102025211048	achmadjunaidi.if@upnjatim.ac.id

Subjects:

T Technology > T Technology (General)

Divisions:

Faculty of Computer Science > Departemen of Informatics

Depositing User:

Mr Faisal Wildan Habibi Habibi

Date Deposited:

22 Jul 2026 06:47

Last Modified:

22 Jul 2026 07:56

URI:

https://repository.upnjatim.ac.id/id/eprint/57221