Claim Missing Document
Check
Articles

Enhancing Natural Disaster Monitoring: A Deep Learning Approach to Social Media Analysis Using Indonesian BERT Variants Fitriani, Karlina Elreine; Faisal, Mohammad Reza; Mazdadi, Muhammad Itqan; Indriani, Fatma; Nugrahadi, Dodon Turianto; Prastya, Septyan Eka
Indonesian Journal of Electronics, Electromedical Engineering, and Medical Informatics Vol. 7 No. 1 (2025): February
Publisher : Jurusan Teknik Elektromedik, Politeknik Kesehatan Kemenkes Surabaya, Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35882/t158qq37

Abstract

Social media has become a primary source of real-time information that can be leveraged by artificial intelligence to identify relevant messages, thereby enhancing disaster management. The rapid dissemination of disaster-related information through social media allows authorities to respond to emergencies more effectively. However, filtering and accurately categorizing these messages remains a challenge due to the vast amount of unstructured data that must be processed efficiently. This study compares the performance of IndoRoBERTa, IndoRoBERTa MLM, IndoDistilBERT, and IndoDistilBERT MLM in classifying social media messages about natural disasters into three categories: eyewitness, non-eyewitness, and don’t know. Additionally, this study analyzes the impact of batch size on model performance to determine the optimal batch size for each type of disaster dataset. The dataset used in this study consists of 1000 messages per category related to natural disasters in the Indonesian language, ensuring sufficient data diversity. The results show that IndoDistilBERT achieved the highest accuracy of 81.22%, followed by IndoDistilBERT MLM at 80.83%, IndoRoBERTa at 79.17%, and IndoRoBERTa MLM at 78.72%. Compared to previous studies, this study demonstrates a significant improvement in classification accuracy and model efficiency, making it more reliable for real-world disaster monitoring. Pre-training with MLM enhances IndoRoBERTa’s sensitivity and IndoDistilBERT’s specificity, allowing both models to better understand context and optimize classification results. Additionally, this study identifies the optimal batch sizes for each disaster dataset: 32 for floods, 128 for earthquakes, and 256 for forest fires, contributing to improved model performance. These findings confirm that this approach significantly improves classification accuracy, supporting the development of machine learning-based early warning systems for disaster management. This study highlights the potential for further model optimization to enhance real-time disaster response and improve public safety measures more effectively and efficiently.
Machine Learning Implementation for Sentiment Analysis on X/Twitter: Case Study of Class Of Champions Event in Indonesia Hafizah, Rini; Saragih, Triando Hamonangan; Muliadi, Muliadi; Indriani, Fatma; Mazdadi, Muhammad Itqan
Indonesian Journal of Electronics, Electromedical Engineering, and Medical Informatics Vol. 7 No. 2 (2025): May
Publisher : Jurusan Teknik Elektromedik, Politeknik Kesehatan Kemenkes Surabaya, Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35882/ijeeemi.v7i2.81

Abstract

Sentiment analysis on social media is becoming an important approach in understanding public opinion towards an event. Twitter, as a microblogging platform, generates a large amount of data that can be utilized for this analysis. This study aims to evaluate and compare the performance of three classification algorithms, namely Support Vector Machine (SVM), Random Forest, and Extreme Gradient Boosting (XGBoost), in sentiment analysis related to the Clash of Champions event in Indonesia. To represent the text data, two feature extraction techniques are used, namely Term Frequency-Inverse Document Frequency (TF-IDF) and Bag of Words (BoW). In addition, Synthetic Minority Over-sampling Technique (SMOTE) is applied to handle data imbalance, while model optimization is performed using GridSearchCV. The research dataset consists of 1,000 tweets collected through web scraping, then manually processed and labeled before model training and testing. The results showed that the TF-IDF technique provided superior results compared to BoW. The Random Forest model with TF-IDF achieved the highest accuracy of 91%, while XGBoost with TF-IDF had the highest Area Under the Curve (AUC) of 0.91. The findings confirm that the selection of appropriate feature extraction techniques and algorithms can improve accuracy in sentiment analysis. This study can be applied in public opinion monitoring and data-driven decision-making. Future research can explore word embedding techniques and transformer-based deep learning models to improve semantic understanding and accuracy of sentiment analysis.
Application of Adaboost Algorithm with SMOTE and Optuna Techniques in Sleep Disorder Classification Anshory, Muhammad Naufal; Mazdadi, Muhammad Itqan; Saragih, Triando Hamonangan; Budiman, Irwan; Saputro, Setyo Wahyu
Indonesian Journal of Electronics, Electromedical Engineering, and Medical Informatics Vol. 7 No. 2 (2025): May
Publisher : Jurusan Teknik Elektromedik, Politeknik Kesehatan Kemenkes Surabaya, Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35882/ijeeemi.v7i2.99

Abstract

Data imbalance is a serious challenge in developing machine learning models for sleep disorder classification. When models are trained on an uneven distribution of classes, classification performance for minority classes such as insomnia and sleep apnea is often low. As a result, the overall accuracy may seem elevated, yet the sensitivity to important cases to be weak. Therefore, this research aims to design and develop a robust sleep disorder classification model with the AdaBoost algorithm, with improved performance through the integration of two main approaches, namely data balancing technique utilizing SMOTE and hyperparameter optimization using Optuna. This research contributes by showing that the combination of the two approaches can significantly improve model performance, not only in terms of global accuracy, but also accuracy on previously overlooked minority classes. The dataset utilized is the Sleep Health and Lifestyle Dataset which consists of 374 synthesized data and is divided into three categories: insomnia, sleep apnea, and none. This method stages include data preprocessing, data division using train-test split (80:20), application of SMOTE to balance the class distribution, hyperparameter tuning using Optuna, and model training with the AdaBoost algorithm. Evaluation was performed using classification metrics: accuracy, precision, recall, and F1-score. Results showed that mix of SMOTE and Optuna yielded the best results, accuracy 90.6%, F1-score 0.83871 for insomnia, and 0.81250 for sleep apnea. This performance was consistently superior to scenarios with no SMOTE or no tuning. This confirms the importance of using combination strategies to obtain fair and accurate classification on medical data. Future research is recommended to use real datasets as well as test the capabilities of this research on other models such as XGBoost or LightGBM.
Revitalisasi Pengemasan Produk UMKM “Woro Production” sebagai Upaya Peningkatan Daya Saing Melalui Penerapan Teknologi Inovatif Mazdadi, Muhammad Itqan; Sari, Anna Khumaira; Normaidah, Normaidah; Saputra, Adryan Maulana; Rahmah, Indah Noor; Ramadhani, Muhammad Irfan; Rahmawati, Nanda Hesti
Jurnal Pengabdian UNDIKMA Vol. 6 No. 4 (2025): November
Publisher : LPPM Universitas Pendidikan Mandalika (UNDIKMA)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33394/jpu.v6i4.17645

Abstract

This community service program aims to strengthen the capacity and technical skills of the “Woro Production” MSME by providing modern packaging equipment and training on its use to improve product efficiency and competitiveness. The implementation method involved training sessions and packaging simulations. Evaluation instruments included observation sheets and interviews to assess the partner’s skills in operating the packaging machine, and the resulting data were analyzed descriptively. The outcomes of this program indicate that participants were able to operate the equipment effectively, and the packaged products demonstrated improved hygiene, practicality, and visual appeal. This initiative is expected to enhance the competitiveness of Woro Production in local, national, and global markets.
KNN-MVO-SMOTE Algorithm for Air Quality Imbalanced Data Classification Rizky, Muhammad Miftahur; Mazdadi, Muhammad Itqan; Muliadi, Muliadi; Faisal, Mohammad Reza; Indriani, Fatma; Rozaq, Hasri Akbar Awal; Yildiz, Oktay
International Journal of Advances in Data and Information Systems Vol. 6 No. 3 (2025): December 2025 - International Journal of Advances in Data and Information Syste
Publisher : Indonesian Scientific Journal

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59395/ijadis.v6i3.1424

Abstract

This research addresses air pollution, a pressing global issue influenced by geographic and temporal factors, using advanced machine-learning techniques to enhance air quality classification. By integrating the K-Nearest Neighbors (KNN) algorithm with the Synthetic Minority Over-sampling Technique (SMOTE) and Multi-Verse Optimization (MVO), we tackle challenges like data imbalance and parameter optimization. Our novel approach, which combines SMOTE and MVO within the KNN framework, has significantly increased classification accuracy to 97%, substantially improving over previous methods. The dataset includes diverse geographic and temporal data, with potential biases acknowledged and addressed. This study highlights the efficacy of merging MVO and SMOTE to optimize classification models, making a substantial contribution to environmental analysis and the fight against air pollution. Future research will explore AutoML technology to improve algorithmic optimization, offering more efficient and adaptive solutions. This pioneering effort emphasizes the critical role of technological innovation in tackling environmental challenges and marks a significant advancement in combating global air pollution.
Peningkatan Akurasi Model Boosting pada Prediksi Kesehatan Tidur Menggunakan Optuna Mazdadi, Muhammad Itqan; Saragih, Triando Hamonangan; Budiman, Irwan; Anshory, Muhammad Naufal
Jurnal Informatika Polinema Vol. 12 No. 2 (2026): Vol. 12 No. 2 (2026)
Publisher : UPT P2M State Polytechnic of Malang

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33795/jip.v12i2.8878

Abstract

Kualitas tidur memiliki peran penting dalam menjaga kesehatan fisik maupun mental, sementara gangguan tidur dapat meningkatkan risiko berbagai penyakit kronis. Perkembangan machine learning membuka peluang untuk melakukan prediksi kesehatan tidur secara lebih akurat melalui pemanfaatan data gaya hidup. Penelitian ini berfokus pada penerapan algoritma boosting, yaitu XGBoost, LightGBM, AdaBoost, dan GradientBoosting, dengan dukungan teknik hyperparameter tuning berbasis Optuna untuk meningkatkan akurasi prediksi. Dataset yang digunakan adalah Sleep Health and Lifestyle Dataset yang memuat variabel demografis, kebiasaan hidup, serta kondisi tidur. Tahapan penelitian meliputi praproses data, pembagian data latih dan uji, pelatihan model, optimasi hyperparameter menggunakan Optuna dengan metode Tree-structured Parzen Estimator (TPE), serta evaluasi model menggunakan metrik akurasi. Hasil eksperimen menunjukkan bahwa tuning dengan Optuna memberikan peningkatan akurasi pada beberapa model, khususnya LightGBM dan AdaBoost, dengan nilai akurasi mencapai 93,3% dan 90,7%. Sementara itu, XGBoost dan GradientBoosting menunjukkan performa stabil dengan akurasi tetap tinggi baik sebelum maupun sesudah tuning. Temuan ini menegaskan bahwa efektivitas tuning bergantung pada karakteristik algoritma yang digunakan. Secara keseluruhan, penelitian ini membuktikan bahwa Optuna dapat menjadi solusi efektif dalam meningkatkan kinerja model boosting untuk prediksi kesehatan tidur. Sebagai arah penelitian lanjutan, disarankan penggunaan metrik evaluasi yang lebih beragam, penerapan teknik penyeimbangan data, serta eksplorasi integrasi dengan metode deep learning untuk memperkaya hasil analisis.
Comparison Between K-Fold Cross Validation And Percentage Split In Decision Tree Algorithms For Anemia Classification Rahmawati, Nanda Putri; Irwan Budiman; Muhammad Itqan Mazdadi; Andi Farmadi; Friska Abadi
Indonesian Journal of Electronics, Electromedical Engineering, and Medical Informatics Vol. 8 No. 1 (2026): February
Publisher : Jurusan Teknik Elektromedik, Politeknik Kesehatan Kemenkes Surabaya, Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35882/ijeeemi.v8i1.315

Abstract

Anemia is a significant global health challenge characterized by a pathological deficit in hemoglobin concentration, often leading to physiological instability. Accurate clinical diagnosis typically relies on complete blood count (CBC) tests, which provide critical hematological parameters for classification. While machine learning models have demonstrated high efficacy in diagnosing anemia, existing research often relies on static data partitioning strategies that may overlook evaluation reliability and performance stability. This study addresses this gap by shifting the focus from architectural benchmarking to validation robustness, specifically evaluating the C4.5 algorithm's performance across different data-splitting techniques. The research uses a dataset comprising 1,281 clinical records with 14 numerical features and 9 anemia-type labels. To assess stability, two distinct partitioning strategies were implemented: a static Percentage Split (ranging from 60:40 to 90:10) and iterative K-Fold Cross Validation (with K values of 3, 5, 7, 10, and 15). Experimental results demonstrate that the C4.5 algorithm achieved its peak performance with the 90:10 Percentage Split, achieving an average accuracy of 99.46%, precision of 98.32%, and recall of 99.28%. In comparison, the K-Fold (K=10) approach yielded a slightly lower but more stable accuracy of 99.19% with a significantly reduced standard deviation (±0.09), highlighting its reliability for clinical applications. While the high-ratio percentage split maximizes training exposure and predictive potential, the K-Fold method provides a more objective, generalizable benchmark by accounting for the entire data distribution. The study further identifies challenges in classifying minority classes, such as Leukemia with thrombocytopenia, due to inherent data scarcity. Ultimately, this research confirms that the C4.5 algorithm, when paired with an optimal partitioning protocol, remains a robust and highly interpretable solution for clinical anemia screening, outperforming several complex modern architectures
The Effect of Smote-Tomek on the Classification of Chronic Diseases Based on Health and Lifestyle Data Muhammad Adika Riswanda; Friska Abadi; Muhammad Itqan Mazdadi; Mohammad Reza Faisal; Rudy Herteno
Indonesian Journal of Electronics, Electromedical Engineering, and Medical Informatics Vol. 8 No. 1 (2026): February
Publisher : Jurusan Teknik Elektromedik, Politeknik Kesehatan Kemenkes Surabaya, Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35882/ijeeemi.v8i1.324

Abstract

Machine learning models for chronic disease prediction are often trained on imbalanced healthcare datasets, where non-disease cases dominate. This condition can lead to misleadingly high accuracy while failing to identify patients with chronic diseases, limiting clinical usefulness. This study aims to analyze the impact of class imbalance on model performance and to evaluate the effectiveness of the SMOTE–Tomek resampling technique in improving chronic disease prediction. This research provides empirical evidence that accuracy alone is insufficient for evaluating healthcare models and demonstrates that imbalance-aware preprocessing is essential for valid and reliable chronic disease detection. Five classification models, such as Support Vector Machine, Random Forest, K-Nearest Neighbors, Gradient Boosting, and XGBoost, were evaluated on a lifestyle-based chronic disease dataset under two conditions: without resampling and with SMOTE–Tomek. Model performance was assessed using accuracy, precision, recall, F1-score, and AUC. Without SMOTE–Tomek, all models failed to detect chronic disease cases, producing near-zero recall and F1-scores despite accuracy exceeding 80%. After applying SMOTE–Tomek, substantial improvements were observed across all models, particularly in recall and AUC. Support Vector Machine achieved the best overall performance, with an accuracy of 92.9%, a precision of 92%, a recall of 93.9%, an F1-score of 0.93, and an AUC of 0.98. The findings confirm that handling class imbalance is a prerequisite for meaningful chronic disease prediction. The consistent increase in recall and AUC across all evaluated models confirms that the improvement stems from enhanced class separability rather than metric inflation. The proposed approach supports more reliable early screening and decision-support systems in preventive healthcare
Comparasion Of Weather Classification Methods On Weather Images Using GLCM Features With Random Forest And Catboost Algoritms Noorhafizi, Muhammad; Saragih, Triando Hamonangan; Mazdadi, Muhammad Itqan; Muliadi, Muliadi; Herteno, Rudy; Rozaq, Hasri Awal Akbar
International Journal of Advances in Data and Information Systems Vol. 7 No. 1 (2026): April 2026 - International Journal of Advances in Data and Information Systems
Publisher : Indonesian Scientific Journal

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59395/ijadis.v7i1.1456

Abstract

Weather image classification is an essential process for improving automated weather information systems. However, most existing studies rely on numerical meteorological data and rarely utilize the textural characteristics embedded in atmospheric imagery. This study addresses that limitation by applying the Gray Level Co-Occurrence Matrix (GLCM) for texture feature extraction combined with Random Forest (RF) and CatBoost algorithms for classification. The dataset, obtained from Kaggle, consists of 1,125 weather images categorized into four classes: cloudy, rain, shine, and sunrise. All images were uniformly normalized and augmented using four rotation angles (0°, 45°, 90°, 135°). GLCM features were extracted with a pixel distance of 1 and gray-level quantization of 8, generating four statistical attributes: contrast, correlation, energy, and homogeneity. Both algorithms were optimized through parameter tuning and evaluated using a 5-fold cross-validation scheme with an 80:20 split ratio. Results show that the Random Forest model (n_estimators = 100, max_depth = 10, random_state = 42) achieved the highest accuracy of 92.43% (±1.12), precision of 92.50%, recall of 92.43%, and F1-score of 92.42%. In comparison, CatBoost (iterations = 100, learning_rate = 0.1, depth = 6) achieved an accuracy of 68.88% (±2.31). The findings demonstrate that GLCM feature extraction combined with Random Forest offers superior stability and accuracy for weather image classification, providing a foundation for efficient and interpretable weather information systems.
Klasifikasi Tanaman Jarak Pagar Menggunakan Algoritme Deep Learning H2O Muhammad Itqan Mazdadi; Rahmat Ramadhani; Triando Hamonangan Saragih; Muhammad Haekal
Jurnal Komputasi Vol. 9 No. 1 (2021)
Publisher : Jurusan Ilmu Komputer Fakultas MIPA Universitas Lampung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.23960/komputasi.v9i1.2774

Abstract

Tanaman jarak pagar merupakan tanaman multi fungsi yang memiliki banyak manfaat dari daun hingga buah. Tanaman jarak pagar sering digunakan untuk produk kecantikan hingga pengganti biodiesel. Penyakit yang menyerang tanaman jarak pagar dapat mengganggu hasil dari tanaman jarak pagar. Kurangnya pakar dibidang ini dan pengetahuan yang dimiliki petani menyebabkan sesuatu yang buruk. Persoalan ini dapat diselesaikan dengan metode Deep Learning. Metode Deep Learning yang digunakan adalah H2O. H2O digunakan karena dapat memberikan hasil komputasi yang cepat dan bisa memberikan akurasi yang baik. Pada penelitian ini bisa kita lihat bahwa H2O memberikan akurasi rata-rata maksimal sebesar 96,066% dengan parameter uji kombinasi data latih dan data uji 60:40, menggunakan satu layer dan jumlah epoch sebanyak 100. Pada penelitian ini membuktikan bahwa H2O bisa digunakan untuk identifikasi penyakit tanaman jarak pagar.
Co-Authors AA Sudharmawan, AA Abdilah, Muhammad Fariz Fata Abdullayev, Vugar Ade Agung Harnawan, Ade Agung Adela Putri Ariyanti Afifa, Ridha Ahdyani, Annisa Salsabila Ahmad Rusadi Ahmad Rusadi Ahmad Rusadi Arrahimi - Universitas Lambung Mangkurat) Ahmad Rusadi Arrahimi - Universitas Lambung Mangkurat) Ahmad Shofi Khairian Ahmad Tajali Aidil Akbar Al Ghifari, Muhammad Akmal Alamudin, Muhammad Faiq Amalia, Raisa Andi - Farmadi Andi Farmadi Andi Farmadi Anna Khumaira Sari Anshory, Muhammad Naufal Ansyari, Muhammad Ridho Antoh, Soterio Ardiansyah Sukma Wijaya Athavale, Vijay Anant Athavale, Vijay Annant budiman, irwan Buih, Putri Helena Junjung Deni Sutaji Dina Arifah Djordi Hadibaya Dodon Turianto Nugrahadi Dwi Kartini Dwi Kartini Dwi Kartini, Dwi Dzira Naufia Jawza Erdi, Muhammad Fatma Indriani Fitriani, Karlina Elreine Fitrinadi Friska Abadi Haekal, Muhammad Hafizah, Rini Hana, Elvina Nur Helma Herlinda Herteno, Rudi Herteno, Rudy Indriani, Fatma Irwan Budiman Irwan Budiman Irwan Budiman Irwan Budiman Irwan Budiman M. Apriannur M. Khairul Rezki Mafazy, Muhammad Meftah Muflih Ihza Rifatama Muhamad Fawwaz Akbar Muhamad Ihsanul Qamil Muhammad Adika Riswanda Muhammad Haekal Muhammad Khairin Nahwan Muhammad Mada Muhammad Mirza Hafiz Yudianto Muhammad Mursyidan Amini Muhammad Reza Faisal, Muhammad Reza Muliadi Muliadi Muliadi Muliadi Muliadi Muliadi Muliadi Muliadi Muliadi Muliadi Muliadi Muliadi Nabella, Putri Noorhafizi, Muhammad Normaidah, Normaidah Nugraha, Muhammad Amir Nursyifa Azizah P., Chandrasekaran Patrick Ringkuangan Prastya, Septyan Eka Putri Nabella Radityo Adi Nugroho Rahmah, Indah Noor Rahmat Hidayat Rahmat Ramadhani Rahmat Ramadhani Rahmawati, Nanda Hesti Rahmawati, Nanda Putri Ramadhani, Muhammad Irfan Ramadhani, Rahmat Ratnapuri, Prima Happy Riadi, Agus Teguh Rifki Izdihar Oktvian Abas Pullah Rifki Rinaldi Rizky, Muhammad Miftahur Rozaq, Hasri Akbar Awal Rozaq, Hasri Awal Akbar Rudy Herteno Saputra, Adryan Maulana Saragih, Triando Hamonangan Satou, Kenji Satrio Yudho Prakoso Setyo Wahyu Saputro Shalehah Siti Fathmah Syahputra, Muhammad Reza Tajali, Ahmad Totok Wianto Wahyu Dwi Styadi Wijaya Kusuma, Arizha Yanche Kurniawan Mangalik YILDIZ, Oktay Yoga Pambudi Yudha Sulistiyo Wibowo Zaini Abdan