Abstract:Abstract: Accurate bone age estimation is essential for monitoring pediatric growth, diagnosing endocrine disorders, and supporting clinical decision-making. Although deep learning has improved prediction accuracy, limited…
ed studies have systematically examined how increasing model depth affects performance and reliability. This study evaluates the effectiveness of progressively deeper convolutional neural networks, specifically EfficientNet variants B0 to B5, for bone age estimation from hand radiographs. Experiments were conducted using 12,611 hand X-ray images from the RSNA Pediatric Bone Age Challenge dataset on Kaggle. To ensure fair comparison, all models were trained using a unified and consistent training pipeline. Model performance was evaluated using Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Concordance Correlation Coefficient (CCC), and Pearson correlation coefficient. The results show a consistent improvement in prediction accuracy as model depth increases. Among the evaluated models, EfficientNet-B5 achieved the best performance, with an MAE of 21.5 months, MAPE of 6.23%, CCC of 0.9148, and Pearson’s r of 0.9203. These findings confirm that model scaling plays a critical role in enhancing prediction robustness and clinical reliability. Future work should emphasize external validation across diverse populations and incorporate interpretability techniques, such as Grad-CAM, to improve clinical transparency and trust.
Keywords: bone age prediction; deep learning; model evaluation; clinical validation
Abstrak: Estimasi usia tulang yang akurat sangat penting untuk memantau pertumbuhan anak, mendiagnosis gangguan endokrin, dan mendukung pengambilan keputusan klinis. Meskipun pembelajaran mendalam telah meningkatkan akurasi prediksi, studi yang secara sistematis meneliti bagaimana peningkatan kedalaman model memengaruhi kinerja dan keandalan masih terbatas. Studi ini mengevaluasi efektivitas jaringan saraf konvolusional yang semakin dalam, khususnya varian EfficientNet B0 hingga B5, untuk estimasi usia tulang dari radiografi tangan. Eksperimen dilakukan menggunakan 12.611 gambar sinar-X tangan dari dataset RSNA Pediatric Bone Age Challenge di Kaggle. Untuk memastikan perbandingan yang adil, semua model dilatih menggunakan alur pelatihan yang terpadu dan konsisten. Kinerja model dievaluasi menggunakan Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Concordance Correlation Coefficient (CCC), dan koefisien korelasi Pearson. Hasil menunjukkan peningkatan yang konsisten dalam akurasi prediksi seiring dengan peningkatan kedalaman model. Di antara model yang dievaluasi, EfficientNet-B5 mencapai kinerja terbaik, dengan MAE sebesar 21,5 bulan, MAPE sebesar 6,23%, CCC sebesar 0,9148, dan Pearson’s r sebesar 0,9203. Temuan ini menegaskan bahwa penskalaan model memainkan peran penting dalam meningkatkan optimasi prediksi dan keandalan klinis. Penelitian selanjutnya dapat menekankan validasi eksternal di berbagai populasi dan menggabungkan teknik interpretasi, seperti Grad-CAM, untuk meningkatkan transparansi dan kepercayaan klinis.
Kata kunci: prediksi usia tulang; deep learning; evaluasi model; validasi klinis
Abstract:Abstract: Conventional semiconductor chip counting at PT Elektronik Indonesia relies on manual weighing, which is prone to human error and inefficiency. This study proposes a desktop-based counting system using a digital…
scanner and image processing. The novelty lies in integrating horizontal-vertical projection with probabilistic Hough transform to robustly detect grid lines, form square cells, and enable accurate unit estimation via average intensity analysis, eliminating the need for reference weighing. Experiments on 15 actual chip images yielded an error rate of 0.009519% and up to 73.674%time efficiency gains compared to the manual method. The system reduces operator dependency, minimizes errors, and accelerates counting, providing a practical machine vision solution for semiconductor production.
Keywords: chip counting; image processing; probabilistic hough transform; grid line detection; time effeciency.
Abstrak: Penghitungan chip semikonduktor konvensional di PT Elektronik Indonesia bergantung pada penimbangan manual, yang rentan terhadap kesalahan manusia dan kurang efisien. Penelitian ini mengusulkan sistem penghitungan berbasis desktop menggunakan scanner digital dan pengolahan citra. Kebaruan terletak pada integrasi proyeksi horizontal-vertikal dengan probabilistic Hough transform untuk mendeteksi garis grid secara kuat, membentuk sel persegi, serta memungkinkan estimasi unit akurat melalui analisis intensitas rata-rata, sehingga menghilangkan kebutuhan penimbangan referensi. Eksperimen pada 15 citra chip aktual menghasilkan tingkat kesalahan 0,009519% dan peningkatan efisiensi waktu hingga 73,674% dibandingkan metode manual. Sistem ini mengurangi ketergantungan operator, meminimalkan kesalahan, dan mempercepat penghitungan, menyediakan solusi machine vision praktis untuk produksi semikonduktor.
Kata kunci: penghitungan chip; pengolahan citra; probabilistic Hough transform; deteksi garis grid; efisiensi waktu.
Abstract:Abstract: One of the main challenges in digital image processing is limited resolution, which makes it difficult to preserve visual details when images are enlarged. Conventional methods such as Bilinear Interpolation are…
e commonly used for image upscaling; however, these approaches often produce blurred images, lose fine textures, and fail to reconstruct complex visual structures. This study aims to enhance digital image resolution by employing a deep learni based approach using a Low-Light Convolutional Neural Network (LLCNN) built upon a Deep Neural Network (DNN) architecture. The dataset used in this study is the DIV2K dataset, which consists of 1,000 high-resolution images. These images were downsampled using scaling factors of ×2, ×3, and ×4 to generate paired Low Resolution–High Resolution (LR–HR) data for training and evaluation. The proposed LLCNN is designed to extract important features such as edges, textures, and local patterns through multiple convolutional layers, followed by non-linear mapping to reconstruct high-resolution images more accurately. Quantitative performance evaluation was conducted using the Peak Signal-to-Noise Ratio (PSNR) and the Structural Similarity Index (SSIM). Model performance was evaluated quantitatively using the Peak Signal-to-Noise Ratio (PSNR) metric. Experimental results showed that the proposed method improved image quality compared to the bilinear method. These results indicate that the deep learning based approach effectively improves image sharpness and structural fidelity, thereby demonstrating its potential for digital image resolution enhancement.
Keywords: deep neural network; image resolution; low-light convolutional neural network; machine learning
Abstrak: Permasalahan utama dalam pengolahan citra digital adalah keterbatasan resolusi yang menyebabkan detail visual sulit dipertahankan ketika citra diperbesar. Metode konvensional seperti Bilinear Interpolation masih banyak digunakan, namun sering menghasilkan citra buram, kehilangan tekstur halus, serta tidak mampu merekonstruksi struktur visual yang kompleks. Penelitian ini bertujuan untuk meningkatkan kualitas resolusi citra digital dengan memanfaatkan pendekatan deep learning berbasis Low-Light Convolutional Neural Network (LLCNN) yang dibangun di atas arsitektur Deep Neural Network (DNN). Data yang digunakan dalam penelitian ini berasal dari dataset DIV2K, yang terdiri dari 1000 citra beresolusi tinggi. Citra tersebut diturunkan menjadi resolusi rendah menggunakan faktor downsampling ×2, ×3, dan ×4 untuk membentuk pasangan data Low Resolution–High Resolution (LR–HR) sebagai data pelatihan dan pengujian. LLCNN dirancang untuk mengekstraksi fitur-fitur penting seperti tepi, tekstur, dan pola lokal melalui beberapa lapisan konvolusi, kemudian melakukan pemetaan non-linear guna merekonstruksi citra resolusi tinggi secara lebih presisi. Evaluasi performa model dilakukan secara kuantitatif menggunakan metrik Peak Signal-to-Noise Ratio (PSNR). Hasil eksperimen menunjukkan bahwa metode yang diusulkan mampu meningkatkan kualitas citra dibandingkan metode bilinear. Hasil ini membuktikan bahwa pendekatan berbasis deep learning efektif dalam meningkatkan ketajaman dan kesesuaian struktur citra digital.
Kata kunci: deep neural network; low-light convolutional neural network; machine learning; resolusi citra
Abstract:Abstract: The capital market plays an important role in describing the economic conditions of a country, and the IHSG is used as the main indicator to observe the movement of all stocks on the Indonesia Stock Exchange. Because…
ecause stock data is volatile and non-linear, the forecasting process becomes challenging, requiring methods that can capture historical patterns more accurately. This study aims to predict IHSG movements using the Long Short-Term Memory (LSTM) model to generate stable short-term predictions. Historical IHSG data was used to train the model, and accuracy was evaluated using Mean Squared Error (MSE). The results show that the model obtained an MSE 6784.0207, RMSE 82.3652 and MAPE 0.88%, indicating a relatively low prediction error rate. The visualization shows that the model's predictions are very close to the actual data, and the 60-day forecasting results show a potential increase in the IHSG of 1.05%. Thus, the LSTM model is capable of providing fairly accurate IHSG predictions and can be a useful tool for investors in analyzing short-term market movements.
Keywords: forecasting; JCI; long short term memory
Abstrak: Pasar modal memiliki peran penting dalam menggambarkan kondisi ekonomi suatu negara, dan IHSG digunakan sebagai indikator utama untuk melihat pergerakan seluruh saham di Bursa Efek Indonesia. Karena data saham bersifat fluktuatif dan tidak linear, proses peramalan menjadi tantangan, sehingga dibutuhkan metode yang mampu menangkap pola historis secara lebih akurat. Penelitian ini bertujuan memprediksi pergerakan IHSG menggunakan model Long Short-Term Memory (LSTM) untuk menghasilkan prediksi jangka pendek yang stabil. Data historis IHSG digunakan untuk melatih model, kemudian akurasi dievaluasi menggunakan Mean Squared Error (MSE). Hasil penelitian menunjukkan bahwa model memperoleh nilai MSE 6784.0207, RMSE 82.3652 dan MAPE 0.88% yang menandakan tingkat kesalahan prediksi relatif rendah. Visualisasi menunjukkan bahwa prediksi model sangat mendekati data aktual, dan hasil forecasting 60 hari ke depan memperlihatkan potensi kenaikan IHSG sebesar 1,05%. Dengan demikian, model LSTM mampu memberikan prediksi IHSG yang cukup akurat dan dapat menjadi alat bantu bagi investor dalam menganalisis pergerakan pasar jangka pendek.
Kata kunci: peramalan; JCI; memori jangka pendek
Abstract:Abstract: The rapid growth of the digital music industry requires accurate music genre classification systems to enhance user experience in streaming services. This study compares a domain-specific Long Short-Term Memory…
(LSTM) network with three Large Language Models (LLMs)—HuBERT, WavLM, and WAV2Vec 2.0—for Music Genre Classification (MGC). The LSTM model was trained using Mel-spectrograms transformed from the GTZAN dataset, while the LLMs were fine-tuned using a smaller set of raw audio samples due to computational constraints. All models were tested on datasets with identical genre labels to ensure a fair evaluation. Results show that the LSTM model achieved the highest accuracy of 97.10%, outperforming HuBERT (86.00%), WavLM (83.00%), and WAV2Vec 2.0 (80.00%). The LSTM demonstrated superior generalization and stability without overfitting, while the LLMs struggled to differentiate between genres with similar acoustic characteristics. These findings indicate that general-purpose pre-trained models, although powerful, are less effective in music-specific tasks due to domain mismatch. Therefore, incorporating music-specific features and architectures remains essential for achieving higher accuracy and reliability in automatic genre classification systems.
Keywords: audio large language models; comparative deep learning; music genre classification.
Abstrak: Pertumbuhan industri musik digital yang pesat menuntut sistem klasifikasi genre musik yang akurat untuk meningkatkan pengalaman pengguna dalam layanan streaming. Penelitian ini dilatarbelakangi oleh perkembangan pesat model pembelajaran mendalam, khususnya jaringan LSTM dan model bahasa berskala besar LLM seperti HuBERT, WavLM, dan WAV2Vec 2.0, yang telah menunjukkan kemampuan representasi audio yang kuat. Tujuan penelitian ini ini membandingkan jaringan Long Short-Term Memory (LSTM) khusus domain dengan tiga model Large Language Models (LLM)—HuBERT, WavLM, dan WAV2Vec 2.0—untuk tugas Klasifikasi Genre Musik (MGC). Metode penelitian melibatkan pelatihan LSTM menggunakan data Mel-spectrogram hasil transformasi dari dataset GTZAN, sementara LLM disesuaikan (fine-tuning) menggunakan data audio mentah dalam jumlah lebih kecil karena keterbatasan komputasi. Seluruh model diuji pada dataset dengan label genre yang sama untuk memastikan evaluasi yang adil. Hasil penelitian menunjukkan bahwa model LSTM mencapai akurasi tertinggi sebesar 97,10%, sedangkan model HuBERT, WavLM, dan WAV2Vec 2.0 masing-masing memperoleh 86,00%, 83,00%, dan 80,00%. Model LSTM menunjukkan kemampuan generalisasi yang lebih baik tanpa overfitting, sedangkan model LLM cenderung kesulitan membedakan genre dengan karakteristik akustik yang mirip. Kesimpulan penelitian ini adalah ketidaksesuaian domain secara signifikan membatasi performa model umum saat diterapkan pada tugas berbasis musik. Oleh karena itu, penggunaan fitur dan arsitektur khusus musik sangat penting dalam membangun sistem klasifikasi genre yang lebih akurat.
Kata kunci: klasifikasi genre musik; model bahasa besar; perbandingan pembelajaran mendalam.
Abstract:Abstract: Soil moisture is an important factor in determining the watering needs of plants for optimal growth. Therefore, accurate monitoring of soil moisture is necessary. This research aims to design and build a soil moisture…
oisture detection tool based on the Decision Tree algorithm with the support of the YL-69 sensor for humidity measurement and the DHT11 sensor for temperature measurement to increase data accuracy. This system uses NodeMCU ESP8266 as a microcontroller and is integrated with an Android application as a user interface. Sensor interpretation data is analyzed using the Decision Tree algorithm to determine soil conditions (dry, damp or wet). The test results show an accuracy level of 95% from 300 data samples. Thus, this system is able to detect soil moisture effectively and can help increase the efficiency of crop management on a household and commercial agricultural scale.
Keywords: agriculture, android, decision tree algorithm, sensors, soil moisture detection
Abstract:Abstract: The rice plant, Oryza sativa, is a major food source in Indonesia. This plant is processed into rice, a staple food for the Indonesian people. Rice growth is crucial to ensure the rice produced is of good quality.…
ty. One part of the rice plant that is susceptible to disease is the leaves, which can inhibit growth and reduce rice quality. Therefore, early detection and accurate classification of rice diseases are crucial to minimize these negative impacts. This has driven the development of a Deep Learning model capable of high-performance automatic classification. This study aims to create a rice leaf classification model using the CNN algorithm and several transfer learning architectures such as ResNet101, VGG16, and Xception. A dataset of 859 rice leaf images collected from the Kaggle website was then processed using augmentation techniques to a total of 2,439 images, plus 215 smartphone photos for external data validation. Thus, the total dataset increased to 2,656 images, covering four categories: leafblast, brownspot, healthy, and hispa. The model was processed in two stages: on the initial dataset (Non-Augmented Dataset) and the Augmented Dataset. The best experimental results were obtained using the ResNet architecture, with a training accuracy of 96.17% and a validation accuracy of 95.22%. Based on the research results, the rice plant disease classification model using deep learning demonstrated good performance.
Keywords: convolutional neural network; deep learning; fine-tuning; image classification; resnet; rice plant
Abstract:Abstract: YouTube is one of the most popular video streaming platforms, but it has constraints that can cause problems when clients have difficulty finding content according to their wishes. The main objective of this study…
udy is to increase user capacity in viewing content specifically in the field of women's empowerment. By using content-based filtering techniques, the system will analyze user preferences and interests through recommendations for women's empowerment content. The data source is via the YouTube API and is analyzed using PHP programming content-based filtering techniques. The system's recommendations provide a list of women's empowerment content with a user request display. The results of the research evaluation obtained a precision value of 62%, meaning that the recommendations match the topic being searched for, namely women's empowerment. The recall value of 84% indicates that the system has succeeded in finding relations from the database. The f1-score value of 72% indicates that there is a balance between precision and recall, meaning that a system is needed that is not only accurate but also complete. While the cosine value shows a score of 0.7071 approaching the maximum value (1.0). The recommendation of the content-based filtering method produces quite effective women's empowerment content.
Keywords: content-based filtering, recommendations, women Empowerment, youtube
Abstract:Abstract: In the era of the Internet of Things (IoT), cyber threats are increasingly complex and dynamic, thus demanding an adaptive and intelligent network security system. This study proposes a Convolutional Neural Network…
work (CNN)-based Intrusion Detection System (IDS) implemented through a Federated Learning (FL) approach in a Non-Independent and Identically Distributed (Non-IID) data environment. This approach allows the model to be trained in a distributed manner across multiple IoT devices without having to collect sensitive data to a central server, thereby maintaining data privacy while increasing the efficiency of the training process. The experiment used the CIC IoT 2023 dataset, which represents various modern IoT network traffic patterns. The results show that the proposed CNN–FL model achieves an overall accuracy of 0.99, with excellent performance in detecting various types of network traffic. The model obtains a perfect recall value (1.00) for normal traffic (Benign), as well as a very high F1-score for DDoS (0.99) and DoS (0.99) attacks. Stable and consistent performance across all five federation rounds demonstrates that this approach is a reliable, efficient, and accurate solution for detecting threats in distributed and privacy-preserving IoT networks.
Keywords: cnn; federated_learning; ids; non-iid; ciciot2023
Abstrak: Dalam era Internet of Things (IoT), ancaman siber semakin kompleks dan dinamis, sehingga menuntut sistem keamanan jaringan yang adaptif dan cerdas. Penelitian ini mengusulkan Intrusion Detection System (IDS) berbasis Convolutional Neural Network (CNN) yang diterapkan melalui pendekatan Federated Learning (FL) pada lingkungan data yang bersifat Non-Independent and Identically Distributed (Non-IID). Pendekatan ini memungkinkan model dilatih secara terdistribusi di berbagai perangkat IoT tanpa harus mengumpulkan data sensitif ke server pusat, sehingga mampu menjaga privasi data sekaligus meningkatkan efisiensi proses pelatihan. Eksperimen menggunakan dataset CIC IoT 2023, yang merepresentasikan berbagai pola lalu lintas jaringan IoT modern. Hasil penelitian menunjukkan bahwa model CNN–FL yang diusulkan mencapai akurasi keseluruhan sebesar 0.99, dengan performa yang sangat baik dalam mendeteksi berbagai jenis lalu lintas jaringan. Model memperoleh nilai recall sempurna (1.00) untuk lalu lintas normal (Benign), serta nilai F1-score yang sangat tinggi untuk serangan DDoS (0.99) dan DoS (0.99). Kinerja yang stabil dan konsisten di seluruh lima putaran federasi membuktikan bahwa pendekatan ini merupakan solusi yang andal, efisien, dan akurat untuk mendeteksi ancaman pada jaringan IoT yang bersifat terdistribusi dan menjaga privasi (privacy-preserving).
Kata kunci: cnn; federated_learning; ids; non-iid; ciciot2023
Abstract:Giving awards is essential to motivate students; however, selecting outstanding students at the junior high school level is often conducted manually and subjectively, which can lead to unfairness and prolonged processing…
time. This study develops a Decision Support System (DSS) that integrates the Analytical Hierarchy Process (AHP) and the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) to support objective and transparent student selection. A quantitative descriptive approach was employed, with data collected through questionnaires, interviews, and documentation at two state junior high schools in Banjarmasin City. Seven assessment criteria were applied: attendance, behavior, uniform neatness, extracurricular participation, academic grades, competition achievements, and disciplinary records. AHP was used to determine the weight of each criterion, while TOPSIS ranked students based on these weights. The web-based system was developed using PHP and MySQL and evaluated using the Technology Acceptance Model (TAM). Results show that academic grades had the highest weight (28.5%), followed by attendance (22.3%) and competition performance (15.2%). The TAM evaluation yielded average scores of 4.32 for Perceived Ease of Use, 4.40 for Perceived Usefulness, 4.15 for Attitudes Towards Use, and 4.28 for Behavioral Intention to Use. The DSS produces accurate rankings, is well-received by users, and offers an efficient, fair, and replicable solution for data-driven educational governance in the digital era.