Article Per Year (5 Year)
p-Index From 2021 - 2026
0.444
P-Index
This Author published in this journals
All Journal Jurnal Manajemen Informatika dan Sistem Informasi Journal of Innovative and Creativity
Nikko Riestian Putra Wardoyo
Universitas PGRI Jombang
Computer Science & IT Education Languange, Linguistic, Communication & Media Mathematics Social Sciences Other

Published : 2 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 2 Documents
Search

 1 
Optimizing Brain Tumor Detection from MRI Images Through Combined VGG16 and ResNet50V2 Models with Batch Normalization Anisah Nabilah; Nikko Riestian Putra Wardoyo
Journal of Innovative and Creativity Vol. 5 No. 3 (2025)
Publisher : Fakultas Ilmu Pendidikan Universitas Pahlawan Tuanku Tambusai

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.31004/joecy.v5i3.3630

Abstract

Brain tumors are one of the most critical and life-threatening health conditions requiring rapid and accurate diagnostic support. Early detection plays a crucial role in determining appropriate medical interventions and improving patient survival rates. With advances in artificial intelligence, particularly computer vision, medical image transmission has emerged as a promising field to address the challenges of manual diagnosis, which is often time-consuming and prone to human error. Magnetic resonance imaging (MRI) is widely used in brain imaging due to its ability to provide detailed structural information, making it an ideal modality for tumor detection and classification. This study employs a Convolutional Neural Network (CNN)-based approach that integrates two deep learning architectures: VGG16 and ResNet50V2, using batch normalization to improve feature extraction and reduce overfitting. Evaluation experiments were conducted on an MRI dataset of 1,311 brain tumor MRI images classified into pituitary, notoma, meningioma, and glioma classes. The aim of this study was to develop a fast, accurate, and efficient method for detecting brain tumors. The results show that the proposed hybrid architecture achieves 98% accuracy, outperforming each pretrained model when applied separately. This study demonstrates that combining multiple CNN architectures with batch normalization can significantly improve the precision and accuracy of brain tumor detection. This approach has the potential to become a valuable diagnostic tool for radiologists, enabling faster and more accurate clinical decision-making. Furthermore, the application of such deep learning models in medical practice could contribute to reducing diagnostic errors and improving patient care in the long term.
KLASIFIKASI TINGKAT STRES PADA DATASET AFFECTNET MENGGUNAKAN HOG, SVM, DAN PSO Nikko Riestian Putra Wardoyo; Bentar Candra Perdana; Anisah Nabilah
Jurnal Manajemen Informatika dan Sistem Informasi Vol. 9 No. 2 (2026): MISI Juni 2026
Publisher : LPPM STMIK Lombok

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.36595/misi.v9i2.2053

Abstract

Pengenalan tingkat stres melalui ekspresi wajah merupakan tantangan besar dalam komputasi afektif akibat deformasi otot yang samar dan dataset yang sangat timpang. Penelitian ini bertujuan untuk mengusulkan dan mengevaluasi arsitektur deteksi stres yang tidak bias menggunakan teknik Undersampling, ekstraksi fitur Histogram of Oriented Gradients (HOG), dan Particle Swarm Optimization (PSO) pada klasifikasi Support Vector Machine (SVM). Dataset AffectNet diseimbangkan melalui Undersampling untuk memaksakan rasio 50:50 yang absolut antara kelas Stres dan Tidak Stres. Fitur HOG diterapkan untuk mengekstraksi orientasi gradien wajah, sementara PSO bertugas mengoptimasi penalti hyperparameter (C) dan nilai gamma dari fungsi kernel SVM RBF. Hasil penelitian menunjukkan bahwa klasifikasi SVM standar menghasilkan jarak baseline yang besar dengan tingkat recall kelas Tidak Stres yang hanya mencapai 1%. Setelah menerapkan teknik Undersampling dan optimasi PSO dengan 30 iterasi, sistem berhasil mendeteksi parameter optimal dan menghasilkan akurasi yang tidak bias (unbiased accuracy) sebesar 59,67%. Meskipun persentase akurasi global tampak menurun dibandingkan dengan akurasi semu (accuracy paradox) yang sering muncul pada data timpang, nilai recall pada kelas minoritas meningkat signifikan hingga 25%, dengan deteksi stres mencapai 96%. Kesimpulannya, arsitektur yang diusulkan berhasil mengeliminasi bias kelas mayoritas, mempercepat waktu ekstraksi fitur menjadi skala milidetik, dan secara efektif mengotomatiskan pencarian parameter SVM, menjadikannya landasan yang kuat untuk sistem pendeteksi psikologis yang berkelanjutan.
Co-Authors Anisah Nabilah Bentar Candra Perdana