SELEKSI FIITUR DAN OPTIMASI HYPERPARAMETER PADA MODEL CATBOOST MENGGUNAKAN PARTICLE SWARM OPTIMIZATION UNTUK KLASIFIKASI TINGKAT HIPERTENSI

Al Afgany, Muhammad Iqbal (2025) SELEKSI FIITUR DAN OPTIMASI HYPERPARAMETER PADA MODEL CATBOOST MENGGUNAKAN PARTICLE SWARM OPTIMIZATION UNTUK KLASIFIKASI TINGKAT HIPERTENSI. Undergraduate thesis, UPN "Veteran" Jawa Timur.

	Text (COVER) 21081010330_Cover.pdf Download (757kB)
	Text (BAB 1) 21081010330_bab1.pdf Download (245kB)
	Text (BAB 2) 21081010330_bab2.pdf Restricted to Repository staff only until 5 December 2027. Download (645kB)
	Text (BAB 3) 21081010330_bab3.pdf Restricted to Repository staff only until 5 December 2027. Download (1MB)
	Text (BAB 4) 21081010330_bab4.pdf Restricted to Repository staff only until 5 December 2027. Download (4MB)
	Text (BAB 5) 21081010330_bab5.pdf Download (232kB)
	Text (DAFTAR PUSTAKA) 21081010330_daftarpustaka.pdf Download (192kB)
	Text (LAMPIRAN) 21081010330_lampiran.pdf Restricted to Repository staff only Download (4MB)

Abstract

This study aims to implement Particle Swarm Optimization (PSO) as a feature selection and hyperparameter optimization method for the CatBoost model to enhance hypertension classification accuracy and determine the effect of PSO implementation on model performance improvement. The research dataset consists of primary data comprising 191 records from Puskesmas Kepatihan Gresik and secondary data comprising 12,500 records from Kaggle, combined into 12,691 records with 11 features. Data splitting was performed with an 80:10:10 ratio for training, validation, and testing. Experiments were conducted by testing 81 scenario combinations of PSO parameters including n_particles (30, 40, 50), w (0.4, 0.73, 0.9), c1 (0.2, 0.5, 1.49), and c2 (1, 1.49, 1.6) to optimize CatBoost hyperparameters such as depth, learning_rate, and l2_leaf_reg. The model was evaluated using accuracy, precision, recall, and F1-score metrics. The test results showed that the CatBoost-PSO model provided significant improvement with accuracy ranging from 95.59%-96.30% (average 95.92%) compared to the baseline model which only achieved 94%. Optimal configuration was obtained at n_particles = 40, c1 = 0.5, c2 = 1, w = 0.4 with the highest accuracy of 96.30%, where PSO successfully reduced features from 11 to 3 main features (Gender, Systolic, Diastolic) with optimal hyperparameters depth = 6, learning_rate = 0.064, and l2_leaf_reg = 1. The n_particles parameter showed the most significant influence on performance with the best computational efficiency at n_particles = 30 (1576.87 seconds), while the c2 parameter demonstrated accuracy consistency across various values despite computational time variations.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Rahajoe, Ani Dijah	NIDN0012057301	anidijah.if@upnjatim.ac.id
Thesis advisor	Wahanani, Henni Endah	NIDN0022097811	henniendah.if@upnjatim.ac.id

Subjects:

T Technology > T Technology (General)

Divisions:

Faculty of Computer Science > Departemen of Informatics

Depositing User:

Muhammad Iqbal Al Afgany

Date Deposited:

05 Dec 2025 08:35

Last Modified:

05 Dec 2025 08:47

URI:

https://repository.upnjatim.ac.id/id/eprint/48039

Actions (login required)

View Item