KLASIFIKASI CITRA MAMOGRAFI MENGGUNAKAN DEEP TRANSFER LEARNING BERDASARKAN FINE-TUNING DAN XGBOOST

Nur Faridah, NIM.: 22206051018 (2024) KLASIFIKASI CITRA MAMOGRAFI MENGGUNAKAN DEEP TRANSFER LEARNING BERDASARKAN FINE-TUNING DAN XGBOOST. Masters thesis, UIN SUNAN KALIJAGA YOGYAKARTA.

Preview	Text (KLASIFIKASI CITRA MAMOGRAFI MENGGUNAKAN DEEP TRANSFER LEARNING BERDASARKAN FINE-TUNING DAN XGBOOST) 22206051018_BAB-I_IV-atau-V_DAFTAR-PUSTAKA.pdf - Published Version Download (2MB) | Preview
	Text (KLASIFIKASI CITRA MAMOGRAFI MENGGUNAKAN DEEP TRANSFER LEARNING BERDASARKAN FINE-TUNING DAN XGBOOST) 22206051018_BAB-II_sampai_SEBELUM-BAB-TERAKHIR.pdf - Published Version Restricted to Registered users only Download (4MB) | Request a copy

Abstract

Breast cancer is a non-skin cancer that usually attacks women. This cancer is the type of cancer with the second most cases after lung cancer, with a mortality rate of 6.8%. Early diagnosis of breast cancer has an important role in the treatment process. One medium for diagnosing breast cancer is using mammography images. This image has a fairly high level of difficulty when interpreted visually, so a computational diagnosis process using deep learning is required. The research aims to offer a precise automatic classification method for breast cancer using deep transfer learning techniques with the CNN VGG-16 architecture as a feature extractor and using two modeling techniques, namely fine-tuning and Extreme Gradient Boosting (XGBoost). The dataset used comes from the Mammographic Image Analysis Society (MIAS) which is used to test the effectiveness of the recommended method for classifying breast cancer into three classes, namely benign, malignant and normal. The test results show that the best accuracy in the fine-tuning technique is 96.02% with the RMSprop optimizer, batch size 64, learning rate 0.01, and epoch 100. Meanwhile, in the XGBoost technique, the accuracy is 93.03% with max depth 8, learning rate 0.4, and num round 500.

Share this knowledge with your friends :

Actions (login required)

View Item