Claim Missing Document
Check
Articles

Found 30 Documents
Search

Brain Tumor Classification with Hybrid Algorithm Convolutional Neural Network-Extreme Learning Machine Radical Rakhman Wahid; Fetty Tri Anggraeni; Budi Nugroho
IJCONSIST JOURNALS Vol 3 No 1 (2021): September
Publisher : International Journal of Computer, Network Security and Information System

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (803.543 KB) | DOI: 10.33005/ijconsist.v3i1.53

Abstract

Brain tumor is a disease that attacks the brains of living things in which brain cells grow abnormally in the area around the brain. Various ways have been done to detect this disease, one of which is through the anatomical approach to medical images. In this study, the authors propose a Convolutional Neural Network (CNN)-Extreme Learning Machine (ELM) hybrid algorithm through Magnetic Resonance Imaging (MRI). ELM was chosen because of its superiority in the training process, which is faster than iterative machine learning algorithms, while CNN was chosen to replace the traditional feature extraction process. The result is CNN-ELM, which has 8 filters in the convolution layer and 6000 nodes in the hidden layer, has the best performance compared to CNN-ELM another model which has different number of filters and number of nodes in the hidden layer. This is evidenced by the average value of precision, recall, and F1-score which is 0.915 while the accuracy of the test is 91.4%.
Performance of Contrast Adjustment in Face Recognition with Training Image under Various Lighting Conditions Budi Nugroho; Eva Yulia Puspaningrum
IJCONSIST JOURNALS Vol 3 No 2 (2022): March
Publisher : International Journal of Computer, Network Security and Information System

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33005/ijconsist.v3i2.63

Abstract

The lighting factor has a very significant effect on facial recognition performance. To reduce the effect of this lighting factor, at the pre-processing stage the researchers used contrast adjustments to the image to improve facial recognition performance. The histogram equalization technique is generally used for contrast adjustment because of its excellent performance to normalize image illumination which is affected by lighting conditions. In this research, empirical experiments were carried out to determine the effect of contrast adjustment using histogram equalization on face recognition in more detail. This research aims to answer the question whether this technique can be used in all image lighting conditions or not. The Robust Regression method is used in this research to recognize faces, which in many cases have very good performance due to lighting factors. Experiments using images in the AR Face Database related to lighting factors. The testing process is carried out by comparing the results of face recognition using the histogram equalization technique in the pre-processing phase and face recognition without pre-processing in each lighting condition. The experimental results show that the use of the histogram equalization technique in pre-processing gives a better face recognition performance effect in low, medium and high lighting conditions. But in very high (extreme) lighting conditions, the use of the histogram equalization technique in pre-processing turns out to have a worse facial recognition performance effect, with an average accuracy of 93.17%, whereas without pre-processing it produces an average accuracy of 94 , 67%.
Performance of Contrast Adjustment Techniques on The Face Recognition Method with Test Data Under Varying Lighting Conditions Budi Nugroho; Hendra Maulana; Anny Yuniarti
IJCONSIST JOURNALS Vol 6 No 2 (2025): March
Publisher : International Journal of Computer, Network Security and Information System

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33005/ijconsist.v6i2.130

Abstract

In the face recognition process influenced by lighting, the application of the image enhancement process at the preprocessing stage plays an important role in normalizing image contrast so that the quality of the input image becomes better. This step is expected to improve face recognition performance. In this study, we implement a lighting-influenced face recognition method, namely Robust Regression, and test several image enhancement techniques in the preprocessing phase to determine their effects on face recognition performance under different image lighting conditions, including Contrast-limited Adaptive Histogram Equalization (CLAHE), Histogram Equalization (Histeq), and Image Intensity Adjustment (Imadjust). HE uses a global technique that adjusts the overall intensity of the image. CLAHE uses a local technique that adjusts the intensity of pixels based on their surrounding areas. Meanwhile, the Imadjust function adjusts the intensity of image pixels based on the specified minimum and maximum values. The experiment is conducted using the AR Face Database which contains images affected by lighting factors. Lighting conditions include several categories, namely low, medium, high, and very high (extreme) lighting conditions. The experimental scenario is carried out by comparing the results of face recognition using several preprocessing techniques on each test data. The experimental results show that image enhancement techniques improve the performance of face recognition. The face recognition approach that adds the CLAHE technique to the preprocessing shows the highest performance of 95.87%. Meanwhile, the face recognition approach that adds the Imadjust technique to the preprocessing shows the lowest performance of 84.38%.
SISTEM KONTROL PRESENTASI REAL-TIME BERBASIS GESTUR TANGAN MENGGUNAKAN METODE LSTM PADA APLIKASI CANVA Salsa Pramudhita Agustiardani; Rizky Parlika; Budi Nugroho
Jurnal Informatika Teknologi dan Sains (Jinteks) Vol 8 No 2 (2026): EDISI 28
Publisher : Program Studi Informatika Universitas Teknologi Sumbawa

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.51401/jinteks.v8i2.7484

Abstract

Penggunaan perangkat konvensional seperti mouse dan keyboard dalam presentasi masih memiliki keterbatasan pada interaksi tanpa sentuhan. Penelitian ini bertujuan untuk mengembangkan sistem kontrol presentasi real-time berbasis gestur tangan menggunakan metode Long Short-Term Memory (LSTM) yang diintegrasikan ke dalam aplikasi Canva. Sistem memanfaatkan MediaPipe untuk mengekstraksi 21 titik landmark tangan sebagai fitur input model klasifikasi. Untuk menjalankan perintah presentasi, terdapat empat kelas gestur yang meliputi gestur one, peace, ok, dan fist. Setiap gestur memiliki fungsi untuk melakukan kontrol presentasi seperti berpindah slide, kembali ke slide sebelumnya, mengaktifkan mode layar penuh, dan keluar dari mode layar penuh. Hasil pengujian menunjukkan model mencapai akurasi sebesar 99%, serta mampu bekerja secara konsisten pada kondisi indoor dan outdoor. Tingkat akurasi yang diperoleh menunjukkan bahwa metode LSTM memiliki performa yang optimal dalam mengenali gestur tangan secara real-time.
PERBANDINGAN METODE GRU DAN XGBOOST DALAM MEMPREDIKSI HARGA PENUTUPAN SAHAM ADRO Eka Maurita Maurita; Rizky Parlika; Budi Nugroho
Jurnal Informatika Teknologi dan Sains (Jinteks) Vol 8 No 2 (2026): EDISI 28
Publisher : Program Studi Informatika Universitas Teknologi Sumbawa

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.51401/jinteks.v8i2.7490

Abstract

Prediksi harga saham merupakan salah satu tantangan yang penting dalam analisis pasar modal sebab perubahan harga saham memiliki sifat fluktuatif serta dipengaruhi oleh berbagai elemen baik eksternal maupun internal. penelitian ini memiliki tujuan untuk membandingkan performa metode Deep Learning dan Machine Learning, yaitu Gated Recurrent Unit (GRU) dan Extreme Gradient Boosting (XGBoost), dalam memprediksi nilai penutupan saham ADRO. Data yang digunakan berupa data time series multivariat yang terdiri dari harga saham ADRO, harga Batubara dunia, dan nilai tukar USD/IDR yang diperoleh dari Yahoo Finance. Tahapan penelitian meliputi pengumpulan dataset, preprocessing data, normalisasi menggunakan Min-Max Normalization, pembentukan sequence dengan teknik sliding window, pelatihan model, denormalisasi, serta evaluasi model menggunakan tiga metrik yaitu MAE, RMSE, dan MAPE. Hasil dari penelitian ini memperlihatkan bahwa model GRU memiliki performa yang lebih baik dibandingkan dengan XGBoost dengan nilai MAE sebesar 50.59, RMSE sebesar 87.85, dan MAPE sebesar 2.12%, sedangkan model XGBoost memperoleh nilai MAE sebesar 66.04, RMSE sebesar 103.47, dan MAPE sebesar 2.60%. berdasarkan hasil tersebut, metode GRU lebih efektif dalam mempelajari pola temporal dan dependensi jangka panjang pada data deret waktu harga saham ADRO sehingga menghasilkan prediksi yang lebih akurat,
PENERAPAN METODE ITEM-BASED COLLABORATIVE FILTERING UNTUK REKOMENDASI MENU PADA KANTIN XYZ Caritta Elizabeth; Rizky Parlika; Budi Nugroho
Jurnal Informatika Teknologi dan Sains (Jinteks) Vol 8 No 2 (2026): EDISI 28
Publisher : Program Studi Informatika Universitas Teknologi Sumbawa

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.51401/jinteks.v8i2.7497

Abstract

Kemajuan teknologi informasi telah mendorong pemanfaatan sistem digital dalam layanan pemesanan makanan, terutama di lingkungan kantin. Jumlah varian menu yang ditawarkan di Kantin XYZ menyulitkan pengguna dalam menentukan pilihan menu yang sesuai dengan selera mereka. Permasalahan tersebut dapat diselesaikan melalui implementasi sistem rekomendasi yang mampu memberikan rekomendasi menu secara personal kepada pengguna. Penelitian ini berfokus pada metode Item-Based Collaborative Filtering (CF) dalam sistem rekomendasi menu Kantin XYZ. Metode tersebut menggunakan data transaksi dan rating pengguna untuk membangun user-item matrix, lalu menghitung antar item menggunakan Cosine Similarity. Setelah itu, prediksi rating dilakukan dengan pendekatan weighted sum berdasarkan nilai similarity antar item. Dataset yang dipakai meliputi data pengguna, data menu, serta data transaksi dan rating. Hasil penelitian mengungkapkan bahwa sistem dapat menghasilkan rekomendasi menu yang relevan dengan berdasarkan pada pola keterkaitan antar item. Evaluasi sistem menggunakan Mean Absolute Error (MAE) menghasilkan nilai sebesar 0,52, yang dinormalisasi terhadap rentang skala rating 1–5 setara dengan sekitar 13% kesalahan relatif. Nilai ini menunjukkan bahwa tingkat kesalahan prediksi cukup rendah, sehingga sistem memiliki kemampuan yang cukup baik dalam menggambarkan preferensi pengguna.
Analisis Pengaruh Peningkatan Kualitas Citra Berbasis Histogram terhadap Kinerja ResNet-18 pada Klasifikasi Karakter Katakana Shella Christanti; Anggraini Puspita Sari; Budi Nugroho
TIN: Terapan Informatika Nusantara Vol 7 No 2 (2026): July 2026
Publisher : Forum Kerjasama Pendidikan Tinggi (FKPT)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47065/tin.v7i2.10078

Abstract

Handwritten Katakana character classification remains a challenging task due to variations in individual writing styles and suboptimal image quality, such as low contrast and unclear character details. These conditions can affect the model's ability to extract discriminative features and accurately recognize characters. This study aims to analyze the impact of histogram-based image enhancement techniques, namely Histogram Equalization (HE), Adaptive Histogram Equalization (AHE), and Contrast Limited Adaptive Histogram Equalization (CLAHE), on the performance of the ResNet-18 model for Katakana character classification. The ETL5 dataset, consisting of 9,200 handwritten character images across 46 character classes, was used in this study. The dataset was divided into training and testing sets with an 80:20 ratio. To increase data diversity and reduce the risk of overfitting, data augmentation using rotation and translation was applied to the training set. The classification process employed a transfer learning-based ResNet-18 model initialized with ImageNet pretrained weights, where only the fully connected layer was fine-tuned using the Adam optimizer for 50 epochs. The experimental results show that CLAHE achieved the best performance with an accuracy of 97.01%, while HE and AHE obtained accuracies of 83.15% and 76.79%, respectively. These findings indicate that CLAHE is more effective than HE and AHE in improving the classification performance of handwritten Katakana characters using ResNet-18, resulting in more accurate predictions. This study contributes by providing a comparative analysis of the effects of three histogram-based image enhancement methods on the performance of ResNet-18 for handwritten Katakana character classification.
Evaluasi Kinerja U-Net ResNet34 dan MDSBN: Studi Komparatif untuk Segmentasi Naskah Kuno Indonesia Rino Zakharia; Budi Nugroho; Eka Prakarsa Mandyartha
Bulletin of Computer Science Research Vol. 6 No. 5 (2026): August 2026
Publisher : Forum Kerjasama Pendidikan Tinggi (FKPT)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47065/bulletincsr.v6i5.1236

Abstract

The digitization of ancient documents is an important step in preserving historical and cultural information. However, the resulting images often suffer from degradation, such as stains, uneven background textures, faded ink, and low contrast, making text-background separation difficult. This study compares two deep learning architectures, namely U-Net ResNet34 and the Modified Deep Semantic Binarization Network (MDSBN), for the segmentation of Indonesian ancient documents. The dataset consists of Balinese palm-leaf manuscripts, Sundanese manuscripts, and additional ancient document images obtained from Wikimedia Commons. The experiments were conducted through a learning rate search and batch size sensitivity analysis, and the models were evaluated using the Dice Coefficient, Intersection over Union (IoU), Precision, Recall, and Root Mean Squared Error (RMSE). This study contributes through a controlled evaluation of both architectures using a consistent dataset, preprocessing pipeline, loss function, evaluation metrics, and computational environment, enabling performance differences to be analyzed more objectively. The results show that U-Net ResNet34 achieved its best performance using a learning rate of 5e-5 and a batch size of 16, with a test Dice score of 0.79338 and a test IoU score of 0.65752. It outperformed MDSBN, which achieved its best performance using a learning rate of 1e-6 and a batch size of 32, with a test Dice score of 0.75338 and a test IoU score of 0.60433. The functional advantage of U-Net ResNet34 is associated with the ability of its residual encoder to extract hierarchical features from complex textures and degradation patterns, while its skip connections help preserve the spatial details of thin text strokes. These characteristics make U-Net ResNet34 more adaptive to variations in degradation within the Indonesian ancient document dataset than the more compact MDSBN architecture.
Optimization of Tea Leaf Disease Detection Based on YOLOv8 Using CBAM and BFP Gilang Rahmadhan Armijantoro; Budi Nugroho; Yisti Vita Via
bit-Tech Vol. 8 No. 3 (2026): bit-Tech
Publisher : Komunitas Dosen Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.32877/bt.v8i3.3394

Abstract

Early identification of tea leaf diseases is essential for sustaining crop productivity and preventing significant yield losses, making accurate automated detection a critical requirement in modern agricultural management. This study aims to improve the robustness of YOLOv8 for disease detection by integrating two complementary optimization modules chosen for their suitability in addressing common challenges in plant imagery: the Convolutional Block Attention Module (CBAM), which enhances discriminative feature focus under complex visual noise, and the Bidirectional Feature Pyramid Network (BiFPN), which strengthens multi-scale feature fusion to capture small or low-contrast lesions. The target diseases include Algal Leaf Spot, Brown Blight, and Grey Blight, using a combined dataset of primary field images and secondary data from Kaggle. Four models were developed—YOLOv8n (baseline), YOLOv8-CBAM, YOLOv8-BiFPN, and YOLOv8-CBAM-BiFPN. Experimental results demonstrate consistent performance gains across all enhanced variants. The baseline model obtained a precision of 0.760, recall of 0.735, and mAP50 of 0.793. Incorporating CBAM increased precision to 0.824 and recall to 0.780, while BiFPN yielded the highest recall of 0.820 with superior multi-scale generalization. The combined CBAM-BiFPN model achieved the strongest overall results, with a precision of 0.879, recall of 0.814, mAP50 of 0.886, and mAP50–90 of 0.739. These findings indicate that integrating CBAM and BiFPN substantially enhances YOLOv8’s capability in complex leaf-disease scenarios and offers practical potential for deployment in real agricultural settings to support faster decision-making and more effective disease management.
Prediksi Harga Bawang Merah di Jawa Timur menggunakan Metode Long Short Term Memory (LSTM) Siti Sri Wahyuni; Budi Nugroho; Chrystya Aji Putra
Prosiding Seminar Nasional Teknologi Informasi dan Bisnis Prosiding Seminar Nasional Teknologi Informasi dan Bisnis (SENATIB) 2026
Publisher : Fakultas Ilmu Komputer Universitas Duta Bangsa Surakarta

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Bawang merah merupakan salah satu komoditas hortikultura strategis yang memiliki peran penting dalam pemenuhan kebutuhan pangan masyarakat Indonesia. Meskipun produksi bawang merah di Provinsi Jawa Timur cenderung meningkat, harga bawang merah masih mengalami fluktuasi yang cukup tinggi sehingga diperlukan metode prediksi yang mampu memodelkan pola data deret waktu secara akurat. Penelitian ini bertujuan menerapkan metode Long Short-Term Memory (LSTM) untuk memprediksi harga harian bawang merah di Provinsi Jawa Timur menggunakan data historis harga periode Januari 2015 hingga Maret 2026 yang diperoleh dari Sistem Informasi Ketersediaan dan Perkembangan Harga Bahan Pokok (SISKAPERBAPO). Model dibangun menggunakan delapan kombinasi parameter yang terdiri atas sequence length (7 dan 30 hari), hidden units (32 dan 64), serta epoch (50 dan 100) kemudian dievaluasi menggunakan MAE, RMSE, MAPE, dan R^2. Hasil penelitian menunjukkan bahwa konfigurasi terbaik diperoleh pada sequence length 7 hari, hidden units 64, dan epoch 100 dengan nilai MAE sebesar 227,91, RMSE sebesar 372,55, MAPE sebesar 0,65%, dan R^2 sebesar 0,9973. Selain itu, hasil prediksi selama 30 hari ke depan menunjukkan kecenderungan penurunan harga secara bertahap. Hasil tersebut menunjukkan bahwa metode LSTM mampu memodelkan pola temporal data harga bawang merah dengan tingkat akurasi yang sangat baik sehingga berpotensi mendukung pengambilan keputusan dalam memantau pergerakan harga komoditas.