Skripsi

EFFICIENTNET-B3 PADA METODE CONVOLUTIONAL NEURAL NETWORK UNTUK KLASIFIKASI CITRA PENYAKIT KULIT

---

Penilaian

---

Early and accurate classification of skin diseases is critical given the high global prevalence WHO estimates nearly 900 million cases worldwide, with dermatitis as the most common and the uneven distribution of dermatological expertise. EfficientNet-B3 within a Convolutional Neural Network (CNN) framework was therefore investigated for multiclass classification of skin disease images. A secondary dataset of 5,000 JPG images sourced from two public Kaggle repositories and comprising five categories (eczema, malignant, melanoma, psoriasis, and seborrheic dermatitis) was preprocessed (resized to 256×256 pixels, normalized, and augmented via rotation, flip, and zoom). Three train-validation-test splits (70:15:15, 60:20:20, and 80:10:10) were evaluated under identical training conditions using sparse categorical cross-entropy loss and the Adam optimizer (learning rate 0.001) for up to 50 epochs with early stopping. Model performance was measured via accuracy, precision, recall, and F1-score on both validation and independent test sets. The 70:15:15 split yielded the most balanced results, with validation accuracy of 93.05% and test accuracy of 91.60% (precision, recall, and F1-score all 91.60%). Although the 80:10:10 split achieved the highest test accuracy (92.00%), its smaller test set introduced potential bias. Confusion matrix analysis highlighted robust inter-class classification with minor confusion among visually similar conditions. Overall, the EfficientNet-B3–based CNN demonstrated reliable feature extraction and consistent classification across five skin disease categories, supporting its potential as a decision-support tool in dermatological practice.

---

Availability

Inventory Code	Barcode	Call Number	Location	Status
2507003565	T176140	T1761402025	Central Library (Reference)	Available but not for loan - Not for Loan

Detail Information

Series Title

-

Call Number

T1761402025

Publisher

Indralaya : Prodi Teknik Informatika, Fakultas Ilmu Komputer Universitas Sriwijaya., 2025

Collation

xiii, 118 hlm.; ilus.; tab.; 29 cm.

Language

Indonesia

ISBN/ISSN

-

Classification

004.07

Content Type

Text

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

Teknik Informatika

Specific Detail Info

-

Statement of Responsibility

MI

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment