EAR DISORDER IMAGE CLASSIFICATION USING EFFICIENTNET-B0 AND XGBOOST

Johan, Ribka Agustina (2026) EAR DISORDER IMAGE CLASSIFICATION USING EFFICIENTNET-B0 AND XGBOOST. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text 22081010345.-cover.pdf Download (851kB)
	Text (Bab 1) 22081010345.-bab1.pdf Download (211kB)
	Text (Bab 2) 22081010345.-bab2.pdf Restricted to Repository staff only until 21 July 2029. Download (806kB) \| Request a copy
	Text (Bab 3) 22081010345.-bab3.pdf Restricted to Repository staff only until 21 July 2029. Download (1MB) \| Request a copy
	Text (Bab 4) 22081010345.-bab4.pdf Restricted to Repository staff only until 21 July 2029. Download (942kB) \| Request a copy
	Text (Bab 5) 22081010345.-bab5.pdf Download (194kB)
	Text (Daftar Pustaka) 22081010345.-daftarpustaka.pdf Download (195kB)

Abstract

Hearing impairment is a growing global health issue that requires early detection to prevent further complications. However, limited access to specialist healthcare services and a shortage of medical experts, particularly in remote areas, make accurate diagnosis challenging. This study aims to develop an automated otoscopic image classification system for ear disorder detection using EfficientNet-B0 as a feature extractor and XGBoost as the classification model. The dataset consists of four classes: normal, earwax plug, myringosclerosis, and chronic otitis media. The proposed method includes image preprocessing consisting of Region of Interest (ROI), resizing, and Contrast Limited Adaptive Histogram Equalization (CLAHE), followed by data augmentation to increase training variability. Features are extracted using a pretrained EfficientNet-B0 model without the top layer, and the resulting feature vectors are classified using XGBoost. Model evaluation is conducted using accuracy, precision, recall, F1-score, classification report, and confusion matrix across six experimental scenarios with different data splits and hyperparameter settings. The results show that the proposed model achieves excellent performance, reaching up to 100% accuracy in the best scenario (90:10 with hyperparameter tuning). In other scenarios, the model consistently achieves high performance with accuracy above 98%. These findings demonstrate that the combination of EfficientNet-B0 and XGBoost is highly effective for otoscopic image classification tasks.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Muttaqin, Faisal	NIDN0030058602	faisalmuttaqin.if@upnjatim.ac.id
Thesis advisor	Puspaningrum, Eva Yulia	NIDN0005078908	evapuspaningrum.if@upnjatim.ac.id

Subjects:

Q Science > QA Mathematics > QA76.87 Neural computers

Divisions:

Faculty of Computer Science > Departemen of Informatics

Depositing User:

Ribka Agustina Johan

Date Deposited:

22 Jul 2026 06:13

Last Modified:

22 Jul 2026 07:42

URI:

https://repository.upnjatim.ac.id/id/eprint/56676