CLASSIFICATION OF ACADEMIC PROCRASTINATION LEVELS AMONG COLLEGE STUDENTS USING A HYBRID XGBOOST AND LIGHTGBM MODEL

Murti, Hapsoro Wisnu (2026) CLASSIFICATION OF ACADEMIC PROCRASTINATION LEVELS AMONG COLLEGE STUDENTS USING A HYBRID XGBOOST AND LIGHTGBM MODEL. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) 22081010073.-cover.pdf Download (1MB)
	Text (Bab 1) 22081010073.-bab1.pdf Download (223kB)
	Text (Bab 2) 22081010073.-bab2.pdf Restricted to Repository staff only until 26 October 2029. Download (780kB)
	Text (Bab 3) 22081010073.-bab3.pdf Restricted to Repository staff only until 26 October 2029. Download (1MB)
	Text (Bab 4) 22081010073.-bab4.pdf Restricted to Repository staff only until 26 October 2029. Download (3MB)
	Text (Bab 5) 22081010073.-bab5.pdf Download (219kB)
	Text (Daftar pustaka) 22081010073.-daftarpustaka.pdf Download (204kB)
	Text (Lampiran) 22081010073.-lampiran.pdf Restricted to Repository staff only until 26 October 2029. Download (18MB)

Abstract

Academic procrastination is the behavior of delaying the completion of academic assignments, which remains prevalent among college students and can hinder the academic process—particularly for seniors working on their theses or final projects—as it has the potential to cause delays in their studies and a decline in academic performance. This study aims to apply and evaluate a hybrid model combining XGBoost and LightGBM to classify students’ levels of academic procrastination into three categories: Low, Moderate, and High. Research data were collected via a questionnaire administered to 394 senior students using the Academic Procrastination Scale–Short Form (APS-S), the Emotion Regulation Questionnaire (ERQ), the Perceived Stress Scale (PSS-10), and demographic data. Labeling was performed based on total APS-S scores using the ideal norm approach. The data processing stages included stratified data splitting with 60:40, 70:30, and 80:20 splits, normalization, encoding, and training data balancing using SMOTE-NC. XGBoost and LightGBM were trained as base models through a search for the optimal parameter combinations; subsequently, the output probabilities of both models were combined using the soft voting method with equal weights. Test results show that the hybrid model’s best performance was achieved in the 70:30 data split scenario, with an accuracy of 0.9328, a macro precision of 0.9322, a macro recall of 0.9238, and a macro F1-score of 0.9264. In this scenario, the hybrid model also improved the classification results for the “Moderate” class compared to both base models. Overall, the combination of XGBoost and LightGBM through the soft-voting hybrid model was able to produce good classification performance, although it did not always achieve the highest performance in every data split scenario.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Muttaqin, Faisal	NIDN0031128503	faisalmuttaqin.if@upnjatim.ac.id
Thesis advisor	Anggraeny, Fetty Tri	NIDN0711028201	fettyanggraeny.if@upnjatim.ac.id

Subjects:

B Philosophy. Psychology. Religion > BF Psychology
Q Science > QA Mathematics > QA76.6 Computer Programming
Q Science > QA Mathematics > QA76.87 Neural computers

Divisions:

Faculty of Computer Science > Departemen of Informatics

Depositing User:

Hapsoro Wisnu Murti

Date Deposited:

27 Oct 2026 04:02

Last Modified:

27 Oct 2026 04:02

URI:

https://repository.upnjatim.ac.id/id/eprint/61178