Claim Missing Document
Check
Articles

Found 29 Documents
Search

Analisis Kinerja Random Forest Dalam Deteksi Gejala Alergi Rongga Mulut Berbasis Warna Gusi Gea, Juli Hartati; Agustinus Rudatyo Himamunanto; Haeny Budiati
Bulletin of Computer Science Research Vol. 5 No. 4 (2025): June 2025
Publisher : Forum Kerjasama Pendidikan Tinggi (FKPT)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47065/bulletincsr.v5i4.657

Abstract

Early detection of allergies in the oral cavity remains challenging due to the subjective nature of visual assessment and limited access to diagnostic facilities. This study proposes a novel approach using the Random Forest algorithm to classify the severity of allergic symptoms based on gum color analysis from digital images. A total of 2,742 gum images were clinically categorized using the Modified Gingival Index (MGI) into mild, moderate, and severe conditions. Preprocessing included conversion to HSV color space and adaptive segmentation using red thresholds on the hue channel (0–10 and 160–180), saturation > 50, and value > 40. Statistical features, including mean, standard deviation, skewness, kurtosis, and entropy, were extracted and normalized using Z-Score. Six parameter combinations were tested with an 80:20 train-test split. The optimal configuration with n_estimators=80, max_depth=9, and min_samples_leaf=2 achieved an accuracy of 95.81%. The highest performance was achieved in the mild class with precision and recall of 98.91%, and stable results in the moderate (93.80%) and severe (94.74%) classes, with only a 0.94% difference. Cross-validation evaluation demonstrated excellent model stability, with an average accuracy of 95.30% and a standard deviation of 0.67%, indicating consistent performance across data subsets. Feature importance analysis showed the dominance of the hue and saturation channels, particularly kurtosis and mean saturation. This study demonstrates that a Random Forest-based allergy detection system using gum color is highly accurate and effective as a non-invasive screening tool in dental and oral health, especially in resource-limited settings, with the potential to improve early screening access in primary healthcare facilities.
Deteksi Penyakit Tanaman Padi (Oryza Sativa L.) Menggunakan Support Vector Machine (SVM) Dan Random Forest Pada Citra Daun Gulo, Bintang Karmila; Agustinus Rudatyo Himamunanto; Jatmika
Bulletin of Computer Science Research Vol. 5 No. 4 (2025): June 2025
Publisher : Forum Kerjasama Pendidikan Tinggi (FKPT)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47065/bulletincsr.v5i4.660

Abstract

Rice (Oryza sativa L.) is a major food crop that is susceptible to disease attacks, which can reduce farmers' productivity and yields. This study aims to develop a digital image-based rice leaf disease classification system using the Support Vector Machine (SVM) and Random Forest algorithms. The dataset consists of three disease classes (Blast, Blight, and Tungro), which are processed through pre-processing stages such as resizing, normalization, and augmentation. Feature extraction is performed using HSV histograms, RGB average values, and Gray Level Co-occurrence Matrix (GLCM) to obtain color and texture characteristics. The data is then divided with a ratio of 80:20 for model training and testing. The evaluation results show that Random Forest provides the best performance with an accuracy of 97.73%, precision and recall values ??above 0.94, and an average F1 score of 0.98. This study shows that a machine learning-based image classification approach can be an effective solution for early detection of diseases in rice plants.
Analisis Beban Kendaraan Terhadap Karakteristik Jalan Menggunakan Metode YOLOv5 Dan Perhitungan ESAL Wanda, Melifan; Himamunanto, Agustinus Rudatyo; Budiati, Haeni
Jutisi : Jurnal Ilmiah Teknik Informatika dan Sistem Informasi Vol 14, No 2: Agustus 2025
Publisher : STMIK Banjarbaru

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35889/jutisi.v14i2.2713

Abstract

Roads are essential infrastructure supporting various types of vehicles; however, excessive loads are a primary cause of surface damage. The increasing volume of vehicles and imbalanced infrastructure development contribute significantly to road deterioration, leading to a reduction in road service life and increased maintenance costs. This study aims to address these issues by developing a system for vehicle detection, classification, and load estimation using the YOLO (You Only Look Once) algorithm a deep learning method capable of detecting and classifying vehicle objects in real time with high speed and accuracy. The data were obtained from CCTV surveillance video recordings. The results indicate that a total of 4,395 vehicles were successfully detected. These detections were then used to estimate the vehicle load using the Equivalent Single Axle Load (ESAL) method. The estimated total daily traffic reached 632,880 vehicles, with a corresponding daily load estimation of 284,214.74 ESAL. The findings highlight the significant impact of vehicle loads on road characteristics and demonstrate the effectiveness of YOLOv5 as a real time tool for monitoring and detecting vehicular load.Keywords: Computer Vision; YOLOv5; Vehicle detection; Vehicle load; Equivalent Single Axle LoadAbstrakJalan merupakan infrastruktur yang penting dalam  menopang berbagai jenis  kendaraan, namun beban berlebih menjadi penyebab utama kerusakan permukaan  pada jalan. Volume kendaraan yang meningkat dan pembangunan infrastruktur yang tidak seimbang  menyebabkan kerusakan pada jalan   yang menyebabkan  pengurangan umur jalan dan meningkatkan biaya perbaikan. Penelitian ini bertujuan untuk mengatasi permasalahan tersebut yaitu dengan membangun Pendeteksi, Klasifikasi dan menghitung  beban kendaraan berbasis Algoritma YOLO (You Only Look Once), sebuah algoritma deep learning yang mampu mendeteksi dan mengklasifikasikan objek kendaraan secara  real-time dengan kecepatan dan akurasi  yang sangat baik. Data yang digunakan diambil dari  rekaman video pengawas CCTV.  Hasil penelitian menunjukan  kendaraan  yang terdeteksi sebanyak 4.395 unit, kendaraan yang  berhasil terdeteksi kemudian dilakukan untuk  perhitungan estimasi beban kendaraan menggunakan perhitungan  Equivalent Single Axle Load (ESAL). Hasil  terhitung dengan total lalu lintas harian mencapai 632.880 unit kendaraan dengan estimasi beban harian sebesar 284.214,74 ESAL. Hasil penelitian menegaskan adanya  pengaruh signifikan beban kendaraan terhadap karakteristik jalan serta menunjukkan efektivitas YOLOv5 sebagai alat dalam memantau  dan mendeteksi beban kendaraan secara  real time 
MODEL PEMBELAJARAN AR INTERAKTIF UNTUK PENGEMBANGAN KOGNITIF ANAK TK DALAM PENGENALAN WARNA DAN BENTUK Tasane, Elshaddai L; Himamunanto, Agustinus Rudatyo; Budiati, Haeni
Jurnal Informatika Teknologi dan Sains (Jinteks) Vol 7 No 2 (2025): EDISI 24
Publisher : Program Studi Informatika Universitas Teknologi Sumbawa

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.51401/jinteks.v7i2.5885

Abstract

Sistem pembelajaran untuk Anak TK masih banyak menggunakan media konvensional seperti buku cetak dan gambar dua dimensi, yang dinilai kurang efektif dalam menyampaikan konsep abstrak seperti warna dan bentuk. Namun, pengenalan warna dan bentuk sangat penting dalam mendukung perkembangan kognitif dan persepsi visual anak. Penelitian ini bertujuan untuk mengembangkan dan menguji aplikasi pembelajaran interaktif berbasis Augmented Reality (AR) yang dirancang untuk memperkenalkan warna dan bentuk secara menarik dan mudah dipahami. Pengembangan aplikasi menggunakan metode Multimedia Development Life Cycle (MDLC), dan dilakukan pengujian sebanyak 30 kali terhadap total 35 objek, yang terdiri dari 24 objek warna dan bentuk 11 objek bentuk. Hasil menunjukkan bahwa seluruh objek warna dikenali dengan akurasi 100%, sedangkan objek bentuk dikenali dengan akurasi 90,9% dengan objek bola gagal dikenali karena pencahayaan dan desain marker. Akurasi keseluruhan sistem sebesar 97,14%. Uji coba terhadap 25 anak usia 4-5 tahun di Tk Kanisius Kadirojo menunjukkan bahwa 88% anak mampu mengoperasikan aplikasi secara mandiri. Penelitian ini memberikan kontribusi dalam pemanfaatan teknologi Marker-Based AR sebagai media pembelajaran inovatif yang mendukung pengembangan kognitif anak Tk secara lebih interaktif dan menyenangkan.
METODE MFCC-SVM UNTUK PENGENALAN TINGKAT EMOSI MANUSIA BERDASARKAN BERAGAM DATASET Nelvina Adonia; Himamunanto, Agustinus Rudatyo; Setyawan, Gogor C.
Jurnal Informatika Teknologi dan Sains (Jinteks) Vol 7 No 2 (2025): EDISI 24
Publisher : Program Studi Informatika Universitas Teknologi Sumbawa

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.51401/jinteks.v7i2.5886

Abstract

Manusia di dalam berbicara pasti memiliki emosi di dalam meluapkan suasana hati tertentu, Namun untuk memahami suasana hati yang dirasakan oleh seseorang yang belum diketahui, suasana hati tersebut yang mempresentasikan adalah sebuah Emosi. Emosi adalah reaksi psikologis dan fisiologis terhadap situasi dan peristiwa yang dirasakan oleh seseorang. Tujuan di dalam penelitian ini yaitu untuk mengklasifikasi dan mengukur emosi seseorang pada suara. Dalam penelitian ini, dirancang sebuah sistem yang mampu mendeteksi atau mengklasifikasi emosi manusia menggunakan sinyal  suaranya. Selain itu, penelitian ini juga memanfaatkan metode Support Vector Machine (SVM) untuk mendeteksi dan mengklasifikasikan suara manusia, dan untuk ekstraksi ciri menggunakan Mel-Frequency, dan untuk mengubah file dari zip menjadi file WAV menggunakan Google Colab. Data suara yang digunakan diambil dari Kaggle seperti RAVDESS, CREMA, dan TORONTO yang berjumlah 12200 data set yang terdiri dari data latih dan data uji. SVM merupakan metode sistem dari machine learning yang digunakan untuk mengklasifikasi suara. Berdasarkan pada penelitian ini, hasil yang dihasilkan klasifikasi emosi melalui suara manusia dengan menggunakan metode SVM ini memiliki akurasi yang cukup tinggi yaitu sebesar 93%.
Implementasi Virtualisasi Proxmox Ve Mengatasi Keterbatasan Sistem Untuk Meningkatkan Efisiensi Layanan Metro Global News Susilo, Zefanya Damar Aristo; Himamunanto, Agustinus Rudatyo; Lase, Kristian Juri Damai
Jurasik (Jurnal Riset Sistem Informasi dan Teknik Informatika) Vol 10, No 2 (2025): Edisi Agustus
Publisher : STIKOM Tunas Bangsa Pematangsiantar

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30645/jurasik.v10i2.914

Abstract

Metro Global News is a media company that, to this day, remains dependent on third-party hosting services as a means to reach its users. This reliance has led to limitations in independently managing server infrastructure, reduced system flexibility, increased operational costs, and minimal control over service configurations. This study aims to implement an internal server infrastructure using virtualization technology based on Proxmox VE as a solution to these challenges. Tools such as Netdata and Portainer were utilized to monitor system performance and resource usage in real-time. The system was built using two virtual machines (VMs), each running separate services within containers, enabling modular management and supervision. Monitoring results through Netdata recorded a total of 4,211 active metrics covering performance indicators such as CPU, RAM, disk I/O, and network usage, with log consumption of only 552 KB at the highest tier. Meanwhile, Portainer reported 4 active containers running application services, 3 storage volumes, and 5 images, with a total RAM allocation of 4.1 GB and the use of 2 CPU cores. All services operated stably under full load conditions without downtime, indicating that the system can function optimally with minimal resources. These findings confirm that Proxmox VE is an effective and cost-efficient solution for building internal IT infrastructure, particularly for medium-scale organizations requiring full control over their systems. This study also opens opportunities for further development through features such as clustering and high availability to enhance service reliability in the future.
Implementasi YOLO11 dan OpenCV untuk Pengenalan Frasa dalam Video Real-Time Bahasa Isyarat Tangan: YOLO11 and OpenCV Implementation for Phrase Recognition in Real-Time Hand Sign Language Videos Swasono, Henoch Yanuar Ari; Himamunanto, Agustinus Rudatyo; Maedjaja, Febe
MALCOM: Indonesian Journal of Machine Learning and Computer Science Vol. 5 No. 3 (2025): MALCOM July 2025
Publisher : Institut Riset dan Publikasi Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.57152/malcom.v5i3.2130

Abstract

Bahasa isyarat adalah alat komunikasi utama bagi para penyandang tunarungu dan tunawicara. Namun, terbatasnya pemahaman bahasa isyarat oleh masyarakat umum sering kali menjadi kendala dalam berkomunikasi. Penelitian ini bertujuan untuk merancang program pengenalan frasa bahasa isyarat tangan BISINDO secara real-time dengan menggunakan algoritma YOLO11 dan library OpenCV. YOLO11 digunakan sebagai metode deep learning untuk mengenali isyarat tangan, sedangkan OpenCV digunakan untuk pemrosesan video real-time dan visualisasi hasil deteksi. Model ini dilatih menggunakan lebih dari 3.000 gambar yang mewakili enam class frasa BISINDO: “saya”, “kamu”, “senang”, “bingung”, “marah”, dan “apa kabar”, sebanyak 263 epoch. Hasil pengujian model menunjukkan rata-rata nilai precision dan recall di atas 0,9; F1-Score sebesar 0,982; mAP50 sebesar 0,993; dan mAP50-95 sebesar 0,938. Pada pengujian real-time, model menunjukkan latency rata-rata stabil di kisaran 80-90ms, frame rate 11-12FPS, dan confidence score rata-rata 0,9 untuk semua class. Berdasarkan Penelitian yang telah dilakukan, disimpulkan bahwa integrasi YOLO11 dan OpenCV berhasil digunakan sebagai algoritma dalam mengenali frasa bahasa isyarat tangan BISINDO secara real-time.
Sistem Pengenalan Citra Dokumen Teks Terdistorsi menjadi Teks Menggunakan Metode Deep Learning Zalukhu, Talenta Teholi; Himamunanto, Agustinus Rudatyo; Budiati, Haeni
Jurnal JTIK (Jurnal Teknologi Informasi dan Komunikasi) Vol 10 No 1 (2026): JANUARY 2026
Publisher : Lembaga Otonom Lembaga Informasi dan Riset Indonesia (KITA INFO dan RISET)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35870/jtik.v10i1.4700

Abstract

A common issue in document image processing is the inability of OCR systems to accurately read text from blurred images. This study aims to develop a deep learning-based OCR pipeline capable of recognizing text in blurred document images. The process begins with image enhancement using the DnCNN model for deblurring, followed by character segmentation and classification of A–Z characters using a CNN trained on the EMNIST Letters dataset. The recognized characters are then reconstructed into complete text. Experiments were conducted on 300 blurred images with varying levels of blur (low, medium, and high). Evaluation using PSNR and SSIM metrics showed improvements in image quality, with an average PSNR of 29,56 dB and SSIM of 0.89. Furthermore, the character classification accuracy reached 95.64%. Compared to the baseline (direct Tesseract OCR without deblurring), the proposed system showed a significant improvement in text readability. These results demonstrate the effectiveness of CNN-based approaches in enhancing OCR performance on blurred document images.
Identifikasi Pola Obyek Kain Tenun Sumba dengan Menggunakan Metode K-Nearest Neighbor (KNN) Budiati, Haeni; Himamunanto, Agustinus Rudatyo; Bolo, Naomi Tena
Upgrade : Jurnal Pendidikan Teknologi Informasi Vol 1 No 1 (2023): Vol. 1 No. 1 Agustus 2023
Publisher : Pendidikan Teknologi Informasi Universitas Bumigora

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30812/upgrade.v1i1.3149

Abstract

Woven fabrics originating from Sumba have their own patterns that distinguish them from other woven fabric patterns throughout Indonesia. The pattern is a distinctive feature that describes the culture of the people in Sumba which is very diverse. To distinguish fabric patterns, one of the algorithms for object recognition is the K-Nearest Neighbor (KNN) algorithm. The KNN algorithm classifies objects based on training data that is closest to the object. Processing works by using metric and eccentricity parameters on training data and input images. This processing will produce text data which is the identification of objects in Sumba woven fabric motifs. Based on the testing that has been done, it successfully identifies the type of object contained in the training data. For types of objects that are not contained in the training data, identification is based on their proximity to the types of objects in the group that contain Sumba woven fabric patterns. The accuracy level of Sumba woven fabric pattern object identification in testing 70 different fabric motif images obtained 62 objects in the input image can be identified correctly (88.57%), while 8 objects in the input image cannot be identified (11.43%).