Garuda - Garba Rujukan Digital

Pendeteksi Penyakit Daun Padi Menggunakan Algoritma YOLOv8 di Desa Jangan-Jangan Kecamatan Pujananting Kabupaten Barru Suandi Aritmawijaya; Fahrim Irhamna Rachman; Rizki Yusliana Bakti
Journal of Muhammadiyah’s Application Technology Vol. 4 No. 3 (2025)
Publisher : Universitas Muhammadiyah Makassar

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.26618/kn1zxt55

ABSTRAKProduksi padi di Indonesia memiliki peran penting dalam menjaga ketahanan pangan nasional, namun produktivitasnya sering mengalami penurunan akibat serangan penyakit pada daun padi. Penyakit seperti blast, bercak coklat, dan hawar daun bakteri merupakan penyakit utama yang dapat menimbulkan kerugian signifikan jika tidak terdeteksi sejak dini. Identifikasi penyakit daun padi secara konvensional umumnya masih dilakukan secara manual dan bergantung pada pengalaman petani, sehingga berpotensi menimbulkan kesalahan diagnosis. Oleh karena itu, penelitian ini bertujuan mengembangkan sistem pendeteksi otomatis penyakit daun padi berbasis deep learning menggunakan algoritma YOLOv8. Dataset diperoleh dari pengambilan citra langsung di lahan pertanian Desa Jangan-Jangan, Kabupaten Barru, yang merepresentasikan kondisi lapangan nyata dan mencakup tiga jenis penyakit utama. Tahapan penelitian meliputi anotasi data menggunakan Roboflow, pelatihan model dengan Google Collab, serta evaluasi performa menggunakan confusion matrix, precision, recall, F1-score, dan mean Average Precision. Hasil pengujian menunjukkan bahwa model YOLOv8 mampu mendeteksi penyakit daun padi dengan akurasi tinggi dan waktu inferensi cepat, sehingga berpotensi diterapkan sebagai solusi deteksi dini penyakit padi secara real-time. Kata Kunci: YOLOv8, Deteksi Penyakit Padi, Deep learning, Citra Digital, Pertanian Presisi, Roboflow,CNN. ABSTRACTRice production in Indonesia plays a crucial role in maintaining national food security, but productivity often declines due to leaf disease attacks. Diseases such as blast, brown spot, and bacterial leaf blight are major diseases that can cause significant losses if not detected early. Conventional rice leaf disease identification is generally still done manually and relies on farmer experience, potentially leading to misdiagnosis. Therefore, this study aims to develop an automatic rice leaf disease detection system based on deep learning using the YOLOv8 algorithm. The dataset was obtained from direct imagery captured in agricultural fields in Jangan-Jangan Village, Barru Regency, which represents real-world conditions and includes three main types of diseases. The research stages include data annotation using Roboflow, model training with Google Colab, and performance evaluation using a confusion matrix, precision, recall, F1-score, and mean Average precision. The test results show that the YOLOv8 model is capable of detecting rice leaf diseases with high accuracy and fast inference time, thus potentially being implemented as a real-time early detection solution for rice diseases. Keyworsds: YOLOv8, Rice Disease Detection, Deep learning, Digital Imagery, Precision Farming, Roboflow,CNN.

Konversi Tulisan Tangan Huruf Kapital Menjadi Teks Menggunakan Metode Deep Learning Berbasis YOLOv8 dan CTC Makmur Jaya Nur; Rizki Yusliana Bakti; Fahrim Irhamna Rachman
Journal of Muhammadiyah’s Application Technology Vol. 4 No. 3 (2025)
Publisher : Universitas Muhammadiyah Makassar

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.26618/9wdk0e43

ABSTRAKPenelitian ini mengkaji pengembangan sistem konversi tulisan tangan ke teks digital menggunakan metode deep learning dengan mengombinasikan arsitektur Convolutional Neural Network (CNN), YOLOv8, dan Connectionist Temporal Classification (CTC). Dataset yang digunakan terdiri dari 700 citra tulisan tangan huruf kapital (A–Z) yang diperoleh dari dokumen resmi Dinas Kependudukan dan Pencatatan Sipil Kabupaten Barru. Tahapan penelitian meliputi prapemrosesan citra berupa grayscale, normalisasi, perataan teks, serta augmentasi data, dilanjutkan dengan anotasi bounding box menggunakan Roboflow. Dataset kemudian dibagi menjadi data pelatihan, validasi, dan pengujian. Model YOLOv8 dilatih untuk mendeteksi karakter dan hasilnya diproses menggunakan CTC untuk menghasilkan teks akhir. Evaluasi menunjukkan performa yang baik dengan precision 98,38%, recall 87,25%, F1-score 92,44%, serta mAP@0.5 sebesar 87,19%. Hasil ini menunjukkan metode yang diusulkan efektif untuk mendukung digitalisasi dokumen administrasi publik.Kata Kunci: YOLOv8, Konversi Tulisan Tangan, Deep Learning, Citra Digital, Administrasi Publik, Roboflow, CNN, CTC ABSTRACTThis study investigates the development of a handwritten text-to-digital text conversion system using deep learning by combining Convolutional Neural Network (CNN), YOLOv8, and Connectionist Temporal Classification (CTC) architectures. The dataset consists of 700 images of uppercase handwritten letters (A–Z) obtained from official documents of the Department of Population and Civil Registration of Barru Regency. The research stages include image preprocessing such as grayscale conversion, normalization, text alignment, and data augmentation, followed by bounding box annotation using Roboflow. The dataset is then divided into training, validation, and testing sets. The YOLOv8 model is trained to detect characters, and the outputs are processed using CTC to generate the final text. Evaluation results demonstrate strong performance, achieving a precision of 98.38%, recall of 87.25%, an F1-score of 92.44%, and an mAP@0.5 of 87.19%. These findings indicate that the proposed method is effective in supporting the digitalization of public administrative documents.Keyworsds: YOLOv8, Handwriting Conversion, Deep Learning, Digital Image, Public Administration, Roboflow, CNN, CTC

Model Autoencoder untuk Deteksi Anomali pada Log Email Mahasiswa Universitas Muhammadiyah Makassar Alvina Damayanti; Fahrim Irhamna Rachman; Darniati Darniati
Journal of Muhammadiyah’s Application Technology Vol. 5 No. 2 (2026)
Publisher : Universitas Muhammadiyah Makassar

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.26618/1wvarw57

ABSTRAK Perkembangan teknologi informasi meningkatkan penggunaan email institusi sebagai sarana komunikasi akademik, namun juga menimbulkan risiko keamanan seperti spam, phishing, dan akses tidak sah. Penelitian ini bertujuan untuk mengimplementasikan model Autoencoder dalam mendeteksi anomali pada log email mahasiswa Universitas Muhammadiyah Makassar. Metode yang digunakan adalah pendekatan kuantitatif berbasis unsupervised learning dengan memanfaatkan data log email yang telah melalui tahap preprocessing dan feature engineering. Model Autoencoder dirancang menggunakan arsitektur encoder-decoder dengan lima hidden layer untuk mempelajari pola aktivitas email normal. Proses deteksi anomali dilakukan menggunakan nilai reconstruction error dengan threshold pada persentil ke-99. Hasil penelitian menunjukkan bahwa model mampu mendeteksi aktivitas anomali dengan baik, di mana sekitar 1% data teridentifikasi sebagai anomali dari keseluruhan dataset. Temuan ini menunjukkan bahwa metode Autoencoder efektif digunakan untuk mendeteksi aktivitas mencurigakan pada sistem email institusi dan berpotensi mendukung peningkatan keamanan sistem informasi di lingkungan perguruan tinggi. Kata Kunci: Autoencoder, deteksi anomali, log email, unsupervised learning, keamanan sistem informasi. ABSTRACT The development of information technology has increased the use of institutional email as a medium for academic communication, but it has also introduced security risks such as spam, phishing, and unauthorized access. This study aims to implement an Autoencoder model for anomaly detection in student email logs at Universitas Muhammadiyah Makassar. The research employed a quantitative approach based on unsupervised learning using email log data that had undergone preprocessing and feature engineering. The Autoencoder model was designed using an encoder-decoder architecture with five hidden layers to learn normal email activity patterns. Anomaly detection was performed using reconstruction error values with a threshold set at the 99th percentile. The results showed that the model was able to detect anomalous activities effectively, where approximately 1% of the data were identified as anomalies from the entire dataset. These findings indicate that the Autoencoder method is effective for detecting suspicious activities in institutional email systems and has the potential to enhance information system security in higher education environments. Keywords: Autoencoder, anomaly detection, email logs, unsupervised learning, information system security.

Penerapan Algoritma KNN dengan K-Fold Cross Validation Untuk Diagnosa Risiko Diabetes Mellitus Hafipa Sudiadarma; Ida Mulyadi; Fahrim Irhamna Rachman
Journal of Muhammadiyah’s Application Technology Vol. 5 No. 2 (2026)
Publisher : Universitas Muhammadiyah Makassar

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.26618/q2zkdm10

ABSTRAK Diabetes mellitus (DM) merupakan penyakit kronis dengan prevalensi yang terus meningkat dan berpotensi menimbulkan berbagai komplikasi serius apabila tidak terdeteksi sejak dini. Keterbatasan metode diagnostik konvensional dalam menangani data kesehatan yang besar dan kompleks mendorong pemanfaatan pendekatan berbasis machine learning. Penelitian ini bertujuan untuk membangun model prediksi risiko diabetes mellitus menggunakan algoritma K-Nearest Neighbor (KNN) dengan metode Stratified K-Fold Cross Validation. Dataset yang digunakan terdiri dari 1.041 data pasien yang diperoleh dari Rumah Sakit Haji Makassar, dengan variabel meliputi usia, tekanan darah, status gula darah sewaktu, indeks massa tubuh, dan lingkar perut. Tahapan penelitian meliputi pemrosesan data, normalisasi menggunakan Standard Scaler, pemodelan KNN dengan metrik jarak Manhattan, serta evaluasi kinerja model menggunakan akurasi, precision, recall, dan F1-score. Hasil penelitian menunjukkan bahwa model KNN mampu mencapai rata-rata akurasi sebesar 87,55% dengan performa yang stabil pada setiap fold. Analisis feature importance menunjukkan bahwa tekanan darah sistolik, lingkar perut, dan gula darah sewaktu merupakan faktor yang paling berpengaruh terhadap status gula darah. Hasil ini menunjukkan bahwa algoritma KNN berpotensi digunakan sebagai alat bantu deteksi dini risiko diabetes mellitus berbasis data kesehatan. ABSTRACT Diabetes mellitus (DM) is a chronic disease with a continuously increasing prevalence and the potential to cause various serious complications if not detected early. The limitations of conventional diagnostic methods in handling large and complex health data have encouraged the use of machine learning-based approaches. This study aims to develop a diabetes mellitus risk prediction model using the K-Nearest Neighbor (KNN) algorithm with the Stratified K-Fold Cross Validation method. The dataset consisted of 1,041 patient records obtained from Haji Hospital Makassar, including variables such as age, blood pressure, random blood glucose level, body mass index, and waist circumference. The research stages included data preprocessing, normalization using Standard Scaler, KNN modeling with the Manhattan distance metric, and model performance evaluation using accuracy, precision, recall, and F1-score. The results showed that the KNN model achieved an average accuracy of 87.55% with stable performance across each fold. Feature importance analysis indicated that systolic blood pressure, waist circumference, and random blood glucose level were the most influential factors affecting blood glucose status. These findings suggest that the KNN algorithm has the potential to be used as a decision-support tool for the early detection of diabetes mellitus risk based on health data..

Title

Found 4 Documents
Search

Abstract

Abstract

Abstract

Abstract

Title Search

Found 4 Documents Search

Abstract

Abstract

Abstract

Abstract

Title

Found 4 Documents
Search