OJA , WIDIYATAMA (2026) PENGEMBANGAN SISTEM COMPUTER-AIDED DETECTION AND DIAGNOSIS (CAD): ANALISIS CITRA KANKER PAYUDARA BERDASARKAN POSSIBILISTIC FUZZY C-MEANS (PFCM) DAN TRANSFER LEARNING MENGGUNAKAN PSEUDO-STADIUM LABELING. FAKULTAS MATEMATIKA DAN ILMU PENGETAHUAN ALAM, UNIVERSITAS LAMPUNG.
|
Text
ABSTRAK.pdf Download (152kB) | Preview |
|
![[img]](http://digilib.unila.ac.id/style/images/fileicons/text.png) |
Text
SKRIPSI FULL.pdf Restricted to Repository staff only Download (8MB) | Request a copy |
|
|
Text
SKRIPSI TANPA BAB PEMBAHASAN.pdf Download (3MB) | Preview |
Abstract
Skin lesion classification based on dermoscopic images is an important approach to support the early detection of skin cancer. However, imbalanced data often becomes a major challenge in multi-class classification, as it can degrade model performance, particularly in recognizing minority classes. This study aims to analyze the effect of data balancing techniques on the performance of a Transformer-based skin lesion classification model using the HAM10000 dataset, which consists of seven skin lesion classes. The model employed is the Swin Transformer with four training scenarios: (1) without any balancing technique, (2) using data augmentation, (3) using class weights, and (4) a combination of data augmentation and class weights.The experimental results show that the model using the combined balancing technique of data augmentation and class weights achieves the best and most balanced performance across evaluation metrics, with an accuracy of 88.77%, precision of 83.28%, sensitivity of 83.42%, specificity of 97.07%, F1-score of 83.10%, and an AUC value of 99.04%. These results indicate that the combination of data variation and penalty through weighting enables the Swin Transformer model to learn the data effectively while minimizing errors compared to training without balancing, with data augmentation only, or with class weights only. Furthermore, this study demonstrates that the choice of data balancing techniques significantly affects model performance, especially for multi-class imbalanced datasets such as HAM10000. Klasifikasi lesi kulit berbasis citra dermatoskopi merupakan pendekatan penting dalam mendukung deteksi dini kanker kulit. Namun, Kata-kata kunci:Kata-kata kunci: Kanker Payudara, Mammografi, Computer-Aided Detection and Diagnosis, Possibilistic Fuzzy C-Means, Transfer Learning.
| Item Type: | Other |
|---|---|
| Subjects: | ?? 000 ?? ?? 001 ?? |
| Divisions: | Fakultas MIPA > Prodi Matematika |
| Depositing User: | 2507275011 Digilib |
| Date Deposited: | 09 Feb 2026 07:21 |
| Last Modified: | 09 Feb 2026 07:21 |
| URI: | http://digilib.unila.ac.id/id/eprint/95929 |
Actions (login required)
 |
View Item |