Klasifikasi Citra Penyakit Paru-Paru Menggunakan Model EfficientNet B2 Dengan Metode Random Sampling

Febrianto, Rengga Yogie (2025) Klasifikasi Citra Penyakit Paru-Paru Menggunakan Model EfficientNet B2 Dengan Metode Random Sampling. Undergraduate thesis, UPN Veteran Jawa Timur.

[img] Text (Cover)
Cover.pdf

Download (1MB)
[img] Text (BAB 1)
BAB 1.pdf

Download (289kB)
[img] Text (BAB 2)
BAB 2.pdf
Restricted to Repository staff only until 24 July 2027.

Download (641kB)
[img] Text (BAB 3)
BAB 3.pdf
Restricted to Repository staff only until 24 July 2027.

Download (612kB)
[img] Text (BAB 4)
BAB 4.pdf
Restricted to Repository staff only until 24 July 2027.

Download (4MB)
[img] Text (BAB 5)
BAB 5.pdf

Download (287kB)
[img] Text (Daftar Pustaka)
DAFTAR PUSTAKA.pdf

Download (231kB)

Abstract

The rapid advancement of deep learning in medical image analysis has revolutionized the diagnosis of lung diseases, with Convolutional Neural Networks (CNNs) such as EfficientNet demonstrating notable success in evaluating chest Xray and CT scan images. Despite its advantages, challenges like class imbalance in medical datasets can lead to suboptimal model performance, making it necessary to explore data sampling techniques. This study evaluates the effectiveness of random sampling methods, including oversampling, original data, and undersampling, in classifying lung diseases using the EfficientNet B2 model. Employing a systematic methodology, this research encompasses data collection, preprocessing, and model training, while using evaluation metrics such as accuracy, precision, recall, and F1-score to assess performance. The results show that the EfficientNet B2 model outperforms the EfficientNet B0 and MobileNet V3 Large model across all scenarios, particularly achieving optimal results with a 70/30 training and testing data split using oversampling, yielding a training accuracy of 99.61% and a testing accuracy of 93.65%. Alternative scenarios using original data and undersampling also proved effective, especially in cases with limited computational resources. Ultimately, while the oversampling strategy stands out in terms of accuracy, the choice of the best methodology should align with specific application needs and available resource constraints to enhance the reliability of medical diagnostic practices.

Item Type: Thesis (Undergraduate)
Contributors:
ContributionContributorsNIDN/NIDKEmail
Thesis advisorDiyasa, I Gede Susrama MasNIDN0019067008igsusrama.if@upnjatim.ac.id
Thesis advisorMandyartha, Eka PrakarsaNIDN0725058805eka_prakarsa.fik@upnjatim.ac.id
Subjects: T Technology > T Technology (General) > T58.6-58.62 Management Information Systems
Divisions: Faculty of Computer Science > Departemen of Informatics
Depositing User: Mr Rengga Yogie Febrianto
Date Deposited: 25 Jul 2025 08:12
Last Modified: 25 Jul 2025 08:12
URI: https://repository.upnjatim.ac.id/id/eprint/40814

Actions (login required)

View Item View Item