Claim Missing Document
Check
Articles

Found 3 Documents
Search

Egg Weight Estimation Based on Image Processing using Mask R-CNN and XGBoost Pardede, Jasman; Rawosi, Muhammad Fadlansyah Zikri Akhiruddin; Setyaningrum, Anisa Putri; Milenio, Rizka Milandga; Chazar, Chalifa
Journal of Applied Data Sciences Vol 6, No 4: December 2025
Publisher : Bright Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47738/jads.v6i4.1004

Abstract

Manually measuring egg weight in the context of livestock and the food industry can pose various problems, including time and labor requirements, the risk of egg damage, consistency and accuracy, and limitations on production scale. To address these issues, an automated egg weight estimation system is essential. This study proposes integrating computer vision and machine learning into a unified workflow that combines segmentation, classification, and regression for practical weight estimation. The proposed pipeline employs Mask R-CNN for egg segmentation, Random Forest (RF) classifier for egg type classification based on color features, and XGBoost for regression using morphological, geometric, color features, and egg type as predictors. The dataset used is 720 images, consisting of 20 eggs (10 chicken and 10 duck), each photographed from 36 rotational angles, and was collected with Ground Truth (GT) weights obtained from a digital scale. Experimental findings show that the RF classifier achieved perfect accuracy (precision, recall, and F1-score = 1.00) in distinguishing chicken and duck eggs. The XGBoost regressor obtained a training performance of MAE = 1.07 g and R² = 0.68, and a validation performance of MAE = 0.23 g and R² = 0.80 under 10-fold grouped cross-validation. Although a Support Vector Regressor baseline reached higher training accuracy (MAE = 0.22 g, R² = 0.96), it failed to generalize on validation (R² 0), highlighting XGBoost’s robustness. The feature importance analysis revealed that there are 4 (four) important features for building an estimation model, namely: Hu moments, eccentricity, elongation, and diagonal length, while color statistics played a complementary role. The novelty of this work lies in combining deep segmentation, color-based classification, and feature-driven regression into a unified framework specifically for egg weight estimation, showing its feasibility as a proof of concept and laying the foundation for future large-scale, calibrated, and externally validated deployment.
UAV Imagery-Based Potential Forest Fire Detection Using YOLOv10 Pardede, Jasman; Pratama, Muhamad Rifki; Milenio, Rizka Milandga
Journal of Applied Data Sciences Vol 7, No 1: January 2026
Publisher : Bright Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47738/jads.v7i1.1052

Abstract

Forest fire mitigation requires an early detection system that is both fast and reliable. This study presents a real-time potential forest fire detection system based on UAV imagery using the YOLOv10 object detection model. The main objective is to enhance the accuracy of detecting fire and smoke in aerial imagery and to minimize false alarms through hyperparameter optimization and data balancing strategies. The dataset used was compiled from Roboflow Universe and Kaggle, consisting of two object classes: fire and smoke, with a slight class imbalance (1329 fire and 1024 smoke). In total, 1,691 annotated images were used, covering various lighting conditions, smoke densities, camera angles, and geographic backgrounds, and were divided into training, validation, and test sets with a ratio of approximately 75:15:10. To address the class imbalance and visual variability, data augmentation techniques such as rotation, flipping, brightness adjustment, and noise addition were applied, along with loss weighting to improve learning performance for the minority smoke class. Model training was conducted using 24 hyperparameter configurations combining six optimizers, two batch sizes, and two learning rates. The best hyperparameters are NAdam optimizer, batch_size 24, and learning_rate 0.001.The best performance of accuracy, precision, recall, F1-score, mean IoU, and mAP were achieved at 0.879, 0.8705, 0.8575, 0.863, 0.7373, and 0.870, respectively. Real-time testing using a DJI Mini 4 Pro UAV with RTMP livestream input demonstrated stable and responsive detection, displaying bounding boxes, class labels, confidence scores, and a “POTENTIAL FOREST FIRE” indicator when both fire and smoke were detected simultaneously. These findings confirm that integrating UAV and YOLOv10 technologies provides an effective and adaptive approach for real-time early detection of potential forest fires.
Machine Learning Optimization on Social Media Sentiment Data for Data Balance Using N-GRAM Milenio, Rizka Milandga; Pardede, Jasman; Kurniasih, Dea
Rekayasa Hijau : Jurnal Teknologi Ramah Lingkungan Vol 10, No 1 (2026)
Publisher : Institut Teknologi Nasional, Bandung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.26760/jrh.v10i1.67-80

Abstract

ABSTRAKKetidakseimbangan kelas merupakan tantangan dalam klasifikasi sentimen pada data media sosial, yang menyebabkan model klasifikasi menjadi bias terhadap kelas mayoritas dan berkinerja buruk pada kelas minoritas. Penelitian ini mengusulkan pendekatan penyeimbangan data berbasis N-Gram untuk mengatasi masalah tersebut dan meningkatkan performa klasifikasi. Tiga model machine learning, yaitu XGBoost, Random Forest, dan Support Vector Machine (SVM), dievaluasi pada dataset yang tidak seimbang maupun seimbang menggunakan akurasi, presisi, recall, dan F1-score sebagai metrik evaluasi. Hasil eksperimen menunjukkan bahwa penyeimbangan data meningkatkan performa semua model tanpa menurunkan kemampuan generalisasi. SVM mencapai performa terbaik pada dataset seimbang dengan akurasi 0,86, presisi 0,87, recall 0,86, dan F1-score 0,86. XGBoost dan Random Forest juga menunjukkan peningkatan performa yang signifikan setelah penyeimbangan, menunjukkan kemampuan yang lebih baik dalam mendeteksi kelas minoritas. Secara keseluruhan, temuan ini menegaskan bahwa pendekatan penyeimbangan data berbasis N-Gram yang diusulkan efektif dalam mengurangi ketidakseimbangan kelas dan meningkatkan ketahanan serta keandalan model klasifikasi sentimen.Kata kunci: klasifikasi sentimen, ketidakseimbangan kelas, n-gram, media sosialABSTRACTClass imbalance is a challenge in sentiment classification of social media data, often causing classification models to be biased toward majority classes and perform poorly on minority classes. This study proposes an N-Gram-based data balancing approach to address this issue and improve classification performance. Three machine learning models, namely XGBoost, Random Forest, and Support Vector Machine (SVM), were evaluated on both imbalanced and balanced datasets using accuracy, precision, recall, and F1-score as evaluation metrics. The experimental results demonstrate that data balancing consistently enhances performance across all models without degrading generalization capability. Among the evaluated methods, SVM achieves the best performance on the balanced dataset, reaching an accuracy of 0.86, precision of 0.87, recall of 0.86, and F1-score of 0.86. XGBoost and Random Forest also show substantial performance improvements after balancing, indicating improved detection of minority sentiment classes. Overall, the findings confirm that the proposed N-Gram-based data balancing approach effectively mitigates class imbalance and improves the robustness and reliability of sentiment classification models.Keywords: Sentiment Classification, Class Imbalance, N-Gram, Social Media