Claim Missing Document
Check
Articles

Found 3 Documents
Search

Implementation of K-Means Clustering Method for Network Traffic Anomaly Detection Haeni Budiati; Antonius Bima Murti Wijaya; Barita Suci Vernando Zebua; Jatmika; Yo’el Pieter Sumihar
Jurnal Mantik Vol. 6 No. 3 (2022): November: Manajemen, Teknologi Informatika dan Komunikasi (Mantik)
Publisher : Institute of Computer Science (IOCS)

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Anomalies may degrade network performance for specific network traffic. Because of its nature, it causes abnormal network traffic. Using the K-means clustering method, this study addresses the formulation of the problem of detecting network bandwidth usage anomalies. The objective of this study is to identify potential network traffic anomalies. This study uses the K-Means Method to predict the value of the network traffic anomalies that will appear. K-Means operates by repeatedly iterating based on the initial cluster entered, until the same cluster results are discovered. The results of the study indicate that predicting the occurrence of anomalies with K-Means will help suppress activities that impede network traffic.
Pemodelan Topik pada Ulasan Google Maps Candi Borobudur Menggunakan Latent Dirchlet Allocation Reonardh Sibarani; Sunneng Sandino Berutu; Kristian Juri Damai Lase; Jatmika Jatmika
Voteteknika (Vocational Teknik Elektronika dan Informatika) Vol 12, No 3 (2024): Vol. 12, No 3, September 2024
Publisher : Universitas Negeri Padang

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24036/voteteknika.v12i3.129028

Abstract

Penelitian ini bertujuan untuk menggali wawasan yang mendalam dari ulasan Google Maps wisatawan tentang Candi Borobudur menggunakan metode Pemodelan Topik Latent Dirichlet Allocation(LDA). Candi Borobudur, salah satu situs warisan dunia yang terkenal di Indonesia, menerima banyak ulasan dari wisatawan yang berbeda-beda. Namun, untuk mengidenfikasi tren, preferensi, dan aspek yang paling penting bagi pengunjung, diperlukan analisis yang lebih dalam. Dengan menggunakan pendekatan LDA, mengelompokkan ulasan-ulasan ini kedalam topik-topik latensial yang memungkinkan untuk mengeksplorasi pola-pola umum dan perbedaan-perbedaan dalam persepsi dan pengalaman pengunjung terhadap Candi borobudur. Proses pemodelan topik ini mencakup beberapa langkah, dimulai dari persiapan data, pra-pemrosesan data, wordcloud, topic coherence serta pemodelan topik. Adapun nilai koherensi paling baik diperoleh dalam model Latent dirichlet allocation (LDA) adalah topik terbaik dengan jumlah 4 topik. Sedangkan dari hasil analisis pemodelan topik pada metode Latent dirichlet allocation (LDA) pada Kata "candi" (0.050883695), "borobudur" (0.02406518), "tiket" (0.018967053), dan "masuk" (0.018507887), menyoroti fokus pada aspek kuil, aksesibilitas, dan biaya masuk. Kata "jalan" (0.012353696) dan "parkir" (0.011362941) menunjukkan perhatian terhadap infrastruktur, sementara "area" (0.01039175) mengacu pada lingkungan sekitar candi. Istilah "rb" (0.009597249) dan "harga" (0.0092093265) merujuk pada biaya dalam rupiah, dan "beli" (0.008586656) terkait dengan pembelian tiket. Hasil ini memberikan wawasan tentang fokus utama dalam pembahasan Borobudur, mencakup kuil, aksesibilitas, infrastruktur, dan biaya.Kata kunci: Google Maps, Pemodelan Topik, Latent Dirichlet Alloation(LDA), Nilai Koherensi This research aims to extract deep insights from tourists' Google Maps reviews of Borobudur Temple using the Latent Dirichlet Allocation (LDA) Topic Modeling method. Borobudur Temple, one of the famous world heritage sites in Indonesia, receives many reviews from different tourists. However, to identify trends, preferences, and aspects that are most important to visitors, a deeper analysis is required. Using the LDA approach, categorizing these reviews into latent topics makes it possible to explore common patterns and differences in visitors' perceptions and experiences of Borobudur Temple. This topic modeling process includes several steps, starting from data preparation, data pre-processing, wordcloud, topic coherence and topic modeling. The best coherence value obtained in the Latent dirichlet allocation (LDA) model is the best topic with a total of 4 topics.. While based on the results of the analysis of topic modeling with the Latent dirichlet allocation (LDA) method on Borobudur reviews get 4 topic models The best coherence value obtained in the Latent dirichlet allocation (LDA) model is the best topic with a total of 4 topics. While from the results of the topic modeling analysis on the Latent dirichlet allocation (LDA) method on the words “temple” (0.050883695), “borobudur” (0.02406518), “ticket” (0.018967053), and “entrance” (0.018507887), highlighting the focus on aspects of the temple, accessibility, and entrance fees. The words “road” (0.012353696) and “parking” (0.011362941) indicate attention to infrastructure, while “area” (0.01039175) refers to the temple's surroundings. The terms “rb” (0.009597249) and “harga” (0.0092093265) refer to costs in rupiah, and “beli” (0.008586656) is related to ticket purchases. These results provide insight into the main focus of the Borobudur discussion, covering temples, accessibility, infrastructure, and costs.Keywords: Google maps, topic modelling, Latent Dirichlet Alloation(LDA), Coherence Score
Analisis Performa Metode YOLOv5-CNN Dalam Meningkatkan Deteksi Dan Pengenalan Ras Kelinci Enjelina Citra Hulu; Agustinus Rudatyo Himamunanto; Jatmika Jatmika
Jutisi : Jurnal Ilmiah Teknik Informatika dan Sistem Informasi Vol 15, No 2 (2026): April 2026
Publisher : STMIK Banjarbaru

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35889/jutisi.v15i2.3499

Abstract

Manual identification of rabbit breeds is time-consuming and error-prone, requiring an automated system based on digital images. This study proposes the YOLOv5-CNN approach to automatically detect rabbit objects and classify their breeds. The first stage uses YOLOv5 to detect rabbits in images and generate bounding boxes. The detected images are then used as input for a Convolutional Neural Network (CNN) model for breed classification. Testing was conducted using a rabbit image dataset divided into 70% training data, 10% validation data, and 20% testing data. In the training and validation stages, the model demonstrated stable learning capabilities in recognizing visual patterns between breeds. Next, testing was conducted on 200 independent test images not used during the training process. The evaluation results showed that the YOLOv5-CNN combination system achieved 96% accuracy on the test data. These findings demonstrate that the integration of object detection and image classification in a single processing pipeline can support automatic rabbit breed identification based on digital images.Keywords: Object detection; Image classification; YOLOv5; EfficientNet-B0; Rabbit breeds AbstrakIdentifikasi ras kelinci secara manual membutuhkan waktu dan rentan kesalahan, sehingga diperlukan sistem otomatis berbasis citra digital. Penelitian ini mengusulkan pendekatan YOLOv5-CNN untuk mendeteksi objek kelinci dan mengklasifikasikan rasnya secara otomatis. Tahap pertama menggunakan YOLOv5 untuk mendeteksi kelinci pada citra dan menghasilkan bounding box, kemudian citra hasil deteksi dijadikan masukan model Convolutional Neural Network (CNN) untuk klasifikasi ras. Pengujian dilakukan menggunakan dataset citra kelinci yang dibagi menjadi 70% data pelatihan, 10% data validasi, dan 20% data pengujian. Pada tahap pelatihan dan validasi, model menunjukkan kemampuan belajar yang stabil dalam mengenali pola visual antar ras. Selanjutnya, pengujian dilakukan pada 200 citra uji independen yang tidak digunakan selama proses pelatihan. Hasil evaluasi menunjukkan bahwa sistem kombinasi YOLOv5–CNN memperoleh akurasi sebesar 96% pada data uji. Temuan ini menunjukkan bahwa integrasi deteksi objek dan klasifikasi citra dalam satu alur pemrosesan dapat mendukung proses identifikasi ras kelinci secara otomatis berbasis citra digital.