Article Per Year (5 Year)
p-Index From 2021 - 2026
1.066
P-Index
This Author published in this journals
All Journal Jurnal Teknologi Informasi dan Multimedia Jurnal Teknik Informatika (JUTIF) Media Jurnal Informatika Bakti Sekawan : Jurnal Pengabdian Masyarakat UPGRADE: Jurnal Pendidikan Teknologi Informasi
Pratama, Gede Yogi
Unknown Affiliation
Religion Humanities Civil Engineering, Building, Construction & Architecture Computer Science & IT Education Health Professions Social Sciences Other

Published : 6 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 6 Documents
Search

 1 
Data Augmentation-Driven Predictive Performance Refinement in Multi-Model Convolutional Neural Network for Cocoa Ripeness Prediction Apriani, Apriani; Switrayana, I Nyoman; Hammad, Rifqi; Irfan, Pahrul; Pratama, Gede Yogi
Jurnal Teknik Informatika (Jutif) Vol. 6 No. 5 (2025): JUTIF Volume 6, Number 5, Oktober 2025
Publisher : Informatika, Universitas Jenderal Soedirman

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.52436/1.jutif.2025.6.5.5298

Abstract

Timely and accurate prediction of cocoa fruit ripeness is critical for optimizing harvest schedules, improving yield quality, and supporting post-harvest processing. Conventional visual inspection methods are prone to subjectivity and inconsistencies, especially when distinguishing among multiple ripeness levels based on fruit age. This study proposes a deep learning approach that leverages multi-model convolutional neural network transfer learning combined with image data augmentation to classify cocoa fruit into four maturity stages derived from fruit age. An augmented dataset of cocoa fruit images was used to fine-tune five well-established pre-trained models: MobileNetV2, Xception, ResNet50, DenseNet121, and DenseNet169. Data augmentation techniques were employed to increase variability and improve model generalization. Model evaluation was conducted using a standard 80:20 training-to-testing split to ensure sufficient data for learning while preserving a representative test set across all ripeness classes. The results demonstrate that DenseNet169 consistently outperformed other models, achieving the highest average accuracy of 85,05%, followed by DenseNet121 84,06%. Across all models, the use of data augmentation led to notable performance gains, highlighting its importance in enhancing predictive capability and reducing overfitting. The proposed framework shows promising potential for automating ripeness classification in agricultural contexts, offering a robust, scalable, and accurate solution for intelligent cocoa harvest management. This work contributes to the growing application of deep learning in precision agriculture, particularly in addressing fine-grained classification problems using limited but enriched visual data.
Rice Leaf Disease Classification Based on ResNet50 and MobileNetV3 Feature Extraction Using Random Forest Pratama, Gede Yogi; Husaini, Rahayun Amrullah; Nasri, Muhammad Haris; Hammad, Rifqi
Media Jurnal Informatika Vol 17, No 2 (2025): Media Jurnal Informatika
Publisher : Teknik Informatika Universitas Suryakancana Cianjur

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35194/mji.v17i2.5939

Abstract

Diseases in rice plants are one of the main factors contributing to decreased agricultural productivity. Early and accurate disease identification is crucial to support effective decision-making in plant disease management. This study aims to compare the performance of deep learning models based on Convolutional Neural Networks (CNN), namely ResNet50 and MobileNetV3, as well as their integration with the Random Forest (RF) algorithm for rice leaf disease classification. The dataset used consists of rice leaf images categorized into several disease classes. Model performance was evaluated using accuracy, precision, recall, and F1-score metrics with a macro-average approach. The results show that the standalone ResNet50 and MobileNetV3 models achieved accuracies of 62.5% and 65.7%, respectively, with macro F1-scores below 0.65, indicating moderate classification performance. However, combining CNN models with Random Forest significantly improved classification performance. The ResNet50 + RF model achieved an accuracy of 99.6%, while the MobileNetV3 + RF model attained the highest accuracy of 99.8%, along with equally high macro-averaged precision, recall, and F1-score values. These findings demonstrate that integrating CNN-extracted features with the Random Forest algorithm enhances the model’s ability to distinguish disease classes more accurately and consistently. Therefore, the hybrid CNN–Random Forest approach shows strong potential as an effective solution for image-based rice plant disease detection systems.
Autism Classification Using MobileNetV3 Feature Extraction and K-Nearest Neighbor Algorithm Husaini, Rahayun Amrullah; Pratama, Gede Yogi; Latif, Kurniadin Abd.; Zulfikri, Muhammad; Augustin, Kartarina
Media Jurnal Informatika Vol 17, No 2 (2025): Media Jurnal Informatika
Publisher : Teknik Informatika Universitas Suryakancana Cianjur

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35194/mji.v17i2.5934

Abstract

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by difficulties in social interaction, communication, and repetitive behaviors. Early detection of ASD is crucial; however, conventional diagnostic methods rely heavily on clinical observation and expert assessment, which can be time-consuming and resource-intensive. Along with the rapid development of artificial intelligence, especially in computer vision and machine learning, automated image-based approaches have gained attention as alternative tools for ASD screening. This study proposes a hybrid classification approach that integrates MobileNetV3 as a feature extraction model with the K-Nearest Neighbor (KNN) algorithm for autism classification using facial image data. Unlike previous CNN–KNN approaches, this study specifically explores the use of MobileNetV3’s lightweight architecture to generate compact and discriminative facial features, which are then classified using KNN to evaluate its effectiveness in low-complexity and resource-efficient settings. This design highlights the novelty of combining an optimized lightweight CNN with a distance-based classifier for autism detection from facial images. The dataset used in this research was obtained from Kaggle and consists of 2,940 labeled facial images of children categorized into Autism and non-Autism classes. This study proposes a hybrid classification approach that combines MobileNetV3 as a lightweight feature extraction model with the K-Nearest Neighbor (KNN) algorithm for autism classification. Experimental evaluations were conducted over multiple independent runs to improve statistical reliability, and model performance was assessed using accuracy, precision, recall, and F1-score. The results indicate that the proposed hybrid model achieves satisfactory and consistent performance while maintaining computational efficiency. These findings suggest that integrating lightweight deep learning models with classical machine learning algorithms can provide an effective and resource-efficient approach for autism classification, with potential applicability as a supportive tool for early ASD screening rather than a definitive clinical diagnosis.
Intervensi Edukasi Digital Marketing untuk Peningkatan Pengetahuan Siswa Siswi Madrasah Aliyah Nasri, Muhammad Haris; Hammad, Rifqi; Husaini, Rahayun Amrullah; Roodhi, Mohammad Najid; Pratama, Gede Yogi
Bakti Sekawan : Jurnal Pengabdian Masyarakat Vol. 6 No. 1 (2026): Juni
Publisher : Puslitbang Sekawan Institute Nusa Tenggara

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35746/bakwan.v6i1.913

Abstract

Advances in digital technology require young people to possess adequate digital literacy skills, particularly in digital marketing, which is now a crucial skill in education, the workplace, and entrepreneurship. However, observations indicate that Madrasah Aliyah (MA) students still lack a grasp of digital marketing concepts and practices. This community service activity aims to improve MA students' knowledge and understanding of basic digital marketing concepts, social media promotion strategies, digital branding principles, and consumer behavior in the digital world. The activity is divided into three stages: preparation, implementation, and evaluation. The preparation phase includes initial discussions with schools and the development of training materials. During the implementation phase, materials are delivered through interactive lectures and discussions, followed by simple practices using digital platforms. Assessments were conducted using pre- and post-tests to measure student knowledge gains. The results showed significant improvement, with an average pre-test score of 42.7 rising to 82.4 in the post-test. This 39.7-point increase indicates that the training successfully strengthened students' understanding of digital marketing concepts. This activity is effective in improving MA students' digital literacy and is relevant to continue with mentoring to better prepare them to face the challenges of the digital era
Pengembangan Platform Intervensi Status Gizi Ibu Hamil Berbasis Integrasi Case-Based Reasoning dan Teori Dempster–Shafer Nasri, Muhammad Haris; Hammad, Rifqi; Pratama, Gede Yogi
Jurnal Teknologi Informasi dan Multimedia Vol. 8 No. 1 (2026): February
Publisher : Sekawan Institut

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35746/jtim.v8i1.905

Abstract

Nutritional problems among toddlers and pregnant women remain a major public health issue in Indonesia, necessitating a decision-support system capable of providing rapid and accurate nu-tritional diagnosis and intervention. This study develops an expert system integrating Case-Based Reasoning (CBR) and the Dempster–Shafer theory to diagnose the nutritional status of toddlers and pregnant women. The CBR method is employed to identify solutions for new cases based on similarity to previous cases, while the Dempster–Shafer theory is utilized to handle un-certainty and combine multiple forms of evidence derived from anthropometric, clinical, and health history parameters. The system was tested using 20 cases involving variables such as body weight, height, mid-upper arm circumference (MUAC), hemoglobin level (Hb), gestational age, and dietary intake. The results indicate that the system achieved an accuracy of 90%, an average confidence level of 82.7%, and a diagnostic precision of 88% when compared to expert nutrition-ists’ assessments. Diagnostic discrepancies occurred in only two cases (10%), both of which ex-hibited parameter values near the classification thresholds. These findings demonstrate that the integration of CBR and the Dempster–Shafer theory enhances the reliability of expert systems in generating accurate and measurable nutritional diagnoses despite data uncertainty, and shows strong potential as a decision-support tool for nutritionists in providing faster, more objective, and evidence-based nutritional interventions.
Augmentasi Data Berbasis GAN dan Ekstraksi Fitur EfficientNetB0 dengan XGBoost untuk Meningkatkan Klasifikasi Penyakit Daun Jagung Anugerah Ikasatya, Ririn; Ikasatya, Ririn Anugerah; Apriliani, Cahya; Firdaus, Fathir Jannatul; Pratama, Gede Yogi
Upgrade : Jurnal Pendidikan Teknologi Informasi Vol 3 No 2 (2026): FEBRUARI
Publisher : Universitas Bumigora

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30812/upgrade.v3i2.6224

Abstract

Corn leaf diseases are one of the main factors contributing to decreased corn productivity. Manual identification of leaf diseases remains subjective, time-consuming, and highly dependent on individual experience. This study aims to improve the performance of image-based corn leaf disease classification through the integration of data balancing techniques, deep feature extraction, and machine learning-based classification methods. The dataset consists of four classes with an imbalanced distribution, namely  Blight with 802 images, Common Rust with 914 images, Gray Leaf Spot with 401 images, and Healthy with 813 images, where GrayLeaf Spot represents the minority class. Data balancing is performed by generating synthetic images using  a convolution-based generative model to increase the number of samples in the minority class. Furthermore, feature extraction is carried out using the EfficientNetB0 architecture, and classification is performed using a gradient boosting-based algorithm. There sults show that the proposed approach improves accuracy from 92.49 percent to 93.29 percent and enhances the model’s ability to recognize the minority class, as indicated by an increase in recall from 69 percent to 78 percent and an improvement in performance balance from 0.76to 0.84. These findings indicate that the proposed method is effective in improving classification performance, particularly for the minority class, without reducing performance on majority classes.
Co-Authors Anugerah Ikasatya, Ririn Apriani Apriani Apriliani, Cahya Augustin, Kartarina Firdaus, Fathir Jannatul Hammad, Rifqi Husaini, Rahayun Amrullah I Nyoman Switrayana Ikasatya, Ririn Anugerah Kurniadin Abd Latif Muhammad Haris Nasri Pahrul Irfan Roodhi, Mohammad Najid Zulfikri, Muhammad