eprintid: 80115 rev_number: 10 eprint_status: archive userid: 12243 dir: disk0/00/08/01/15 datestamp: 2026-11-20 06:33:28 lastmod: 2026-11-20 06:33:28 status_changed: 2026-11-20 06:33:28 type: thesis metadata_visibility: show contact_email: muchti.nurhidaya@uin-suka.ac.id creators_name: Nafal Maulana Abdillah, NIM.: 22106010055 title: PEMODELAN MULTILAYER PERCEPTRON UNTUK KLASIFIKASI STATUS KEKAMBUHAN KANKER PARU BERDASARKAN PROFIL METILASI DNA (STUDI KASUS : PASIEN LUNG ADENOCARCINOMA STADIUM I PASCAOPERASI) ispublished: pub subjects: 560 divisions: jur_mat full_text_status: restricted keywords: Kanker Paru-Paru; Metilasi DNA; High-Dimensional Data; Multilayer Perceptron; NumPy; SHAP note: Dr. Epha Diana Supandi, S.Si., M.Sc. abstract: Although the surgical success rate for Stage I Lung Adenocarcinoma (LUAD) is relatively high, approximately 27% of patients still experience fatal post-operative recurrence. DNA methylation profiles serve as potential epigenetic predictors for detecting this risk; however, their analysis is often constrained by high feature dimensionality. This study aims to classify recurrence status by constructing a Multilayer Perceptron (MLP) algorithm from scratch using NumPy to provide mathematical transparency and optimization that enhances computational efficiency. The optimal model was achieved using Variance Threshold (0.001) and Lasso (maximum of 5,000 features), followed by MLP training with an architecture of 256 and 64 hidden neurons, a learning rate of 0.1, and 5,000 epochs. Evaluation of the best model on the test dataset yielded an Accuracy of 65% and a Recall of 43%, alongside a Cross-Validation Accuracy of 63.04% and Recall of 56.67%, where predictive performance was influenced by class imbalance. Nevertheless, evaluation using SHapley Additive exPlanations (SHAP) paired with the hg38 reference genomic annotation successfully identified the most dominant biomarker genes influencing tumor recurrence: TTC6, C5orf66, TDRD10, and BHLHA15. Overall, this modeling approach not only classifies recurrence status but also successfully delivers mathematical transparency and biological relevance underlying tumor recurrence. date: 2026-08-13 date_type: published pages: 163 institution: UIN SUNAN KALIJAGA YOGYAKARTA department: FAKULTAS ILMU TARBIYAH DAN KEGURUAN thesis_type: skripsi thesis_name: other citation: Nafal Maulana Abdillah, NIM.: 22106010055 (2026) PEMODELAN MULTILAYER PERCEPTRON UNTUK KLASIFIKASI STATUS KEKAMBUHAN KANKER PARU BERDASARKAN PROFIL METILASI DNA (STUDI KASUS : PASIEN LUNG ADENOCARCINOMA STADIUM I PASCAOPERASI). Skripsi thesis, UIN SUNAN KALIJAGA YOGYAKARTA. document_url: https://digilib.uin-suka.ac.id/id/eprint/80115/1/22106010055_BAB-I_IV-atau-V_DAFTAR-PUSTAKA.pdf document_url: https://digilib.uin-suka.ac.id/id/eprint/80115/2/22106010055_BAB-II_sampai_SEBELUM-BAB-TERAKHIR.pdf