Journal of Informatics And Telecommunication Engineering
Not a member yet
373 research outputs found
Sort by
Comparative Analysis of Naïve Bayes and K-Nearest Neighbor for Lexicon-Based Emotion Classification of Paxel App User Reviews
No full textThe rapid growth of app-based delivery services has increased the importance of understanding user emotions as an indicator of service quality. User reviews on digital platforms provide valuable insights into customer perceptions, satisfaction levels, and service-related issues. This study aims to compare the performance of Naïve Bayes and K-Nearest Neighbor (KNN) algorithms in classifying user emotions related to the Paxel application. The dataset was collected from Google Play Store and X (Twitter) using web scraping techniques and subsequently processed through text pre-processing stages, including case folding, tokenization, and stopword removal. Emotion labels were assigned using the NRC Indonesian Emotion Lexicon, while feature extraction was performed using the TF-IDF method. To address class imbalance, the Synthetic Minority Oversampling Technique (SMOTE) was applied prior to model training. Experimental results show that the Naïve Bayes model achieved the highest overall accuracy of 90.83% with a weighted F1-score of 0.90, while the KNN model obtained an accuracy of 81.21% and a weighted F1-score of 0.77. Both models performed well in identifying happy, sad, and neutral emotions, whereas anger remained the most challenging class to classify. Overall, Naïve Bayes demonstrated more consistent and reliable performance for sentiment analysis tasks.
Performance Evaluation of Augmented Reality-Based Smart Farming for Rice and Corn Pest Detection
No full textAn Augmented Reality application for detecting pests and diseases in rice and corn has been developed to overcome the limitations of visual identification in the field, which still relies on subjective interpretation by users. This system utilises image processing and AR overlay based on smart farming to classify symptoms in real time, improving the precision of diagnosis and consistency in control decision-making. This study aims to design, implement, and evaluate the performance of an augmented reality (AR)-based smart farming system for the visual and interactive detection of pests and diseases in rice and corn crops. The research method uses an evaluative approach by assessing the performance of the Augmented Reality system in the field based on detection accuracy, operational reliability, and the suitability of the results to the predetermined performance indicators. Testing was conducted in Gampong Releut Barat, Dewantara District, North Aceh. The results showed that pest and disease detection accuracy increased from 42.4% to 66.7%, with a system response time of <2 seconds, accompanied by an 18% reduction in crop damage and a 24% increase in productivity, confirming the reliability of the system for field diagnosis. This achievement is significant because it meets the operational performance threshold for smart farming and demonstrates the system's readiness for adoption as an Augmented Reality-based decision support tool at the farmer level. The research conclusion indicates that Augmented Reality-based smart farming has the potential to improve detection accuracy, control efficiency, and crop productivity as a support for precision agriculture and sustainable village food security
Decision Support System for Human Resource Program Prioritization Using AHP–SMART
No full textHuman Resource (HR) programs are a crucial aspect of improving the quality of life in rural communities; however, they are often constrained by limited resources and urgent needs. Therefore, this study implements the Analytical Hierarchy Process (AHP) and Simple Multi-Attribute Rating Technique (SMART) methods within a decision support system. The AHP method is employed to determine the criteria weights based on the relative importance of various factors influencing program implementation, including cost, benefits, community participation, needs, sustainability, ease of implementation, and risk of failure. Subsequently, the SMART method is applied to rank program alternatives based on the evaluated criteria. Accuracy testing shows that the system produces results fully consistent with manual calculations, achieving an accuracy rate of 100%. Functional testing using the black-box method indicates that all system features operate properly without errors. Meanwhile, the User Acceptance Test (UAT) results demonstrate that all respondents provided positive evaluations (scores ranging from 3 to 5), with no reported dissatisfaction, indicating that the system is feasible and well accepted by users. The results reveal that the integration of AHP and SMART provides accurate program priority recommendations, with the Community Health Program (0.7150) ranked as the top priority and Food Security Training (0.6550) as the second priority. This decision support system is expected to enhance the efficiency and accuracy of human resource decision-making in Blang Pulo Villag
Digital Marketing and Consumer Engagement for Traditional Handwoven Products through a Web-Based Retail System
No full textTraditional handwoven products often struggle to reach wider markets due to the continued reliance on conventional selling practices, which limits visibility and reduces opportunities for consumer engagement in digital retail environments. This study aims to develop and evaluate a web-based retail platform that supports the digital marketing of traditional handwoven products and enhances consumer interaction. A Research and Development approach was employed, integrating essential retail features such as product catalog management, ordering processes, user administration, and interaction interfaces. The platform was evaluated through expert validation to assess content relevance, functional testing using automated tools to ensure system accuracy, performance assessment to examine system efficiency, and usability testing involving 25 users to measure ease of use and consumer acceptance. The results demonstrate that the platform performs reliably, achieves complete functional accuracy, and exhibits high usability, indicating strong approval among users. Performance testing also shows that the platform operates efficiently and is suitable for real retail use. The implications of this study suggest that web-based retail platforms can strengthen digital marketing practices, expand market accessibility for traditional artisans, and contribute to improved consumer experience within craft-based retail sectors
Comparative Study of VGG16 and MobileNet Architectures for Rice Leaf Disease Classification Using CNN
No full textRice is a primary commodity in Indonesia's agricultural sector, playing a vital role in national food security. However, rice productivity is frequently disrupted by leaf diseases such as Bacterial Leaf Blight, Brown Spot, Leaf Blast, and Narrow Brown Spot. This study aims to develop an automated rice leaf disease identification model using the Convolutional Neural Network (CNN) method with a transfer learning approach. Two CNN architectures, VGG16 and MobileNet, were trained using a dataset of 2,190 rice leaf images divided into five classes. The research process included data collection, preprocessing, model training, and performance evaluation using a confusion matrix. The results show that the VGG16 model achieved an accuracy of 98%, while MobileNet reached 95% accuracy. Thus, this method can serve as a modern solution for identifying rice plant diseases, supporting early detection efforts and enhancing agricultural productivity
Optimization of Feature Extraction in Images Using Variants of Decomposition Algorithms
Full text linkThis research aims to optimize the feature extraction process in digital images using two decomposition algorithms, namely Haar and Riyad. Feature extraction is an important step in digital image processing, used to extract significant information from images for applications such as pattern recognition, medical image analysis, and surveillance systems. Haar and Riyad algorithms are tested on three types of images: grayscale, color, and texture. Results show that Haar's algorithm excels in processing speed with an average time of 121.67 ms, making it ideal for real-time applications. In contrast, the Riyad algorithm showed higher feature detection accuracy, achieving an average of 93.33% on complex images, despite requiring a longer processing time of 154 ms. This research shows that the selection of a feature extraction algorithm should consider the type of image and the application needs. Haar's algorithm is suitable for real-time surveillance applications, while Riyad is more suitable for in-depth analysis such as on medical images. The significant contribution of this research is that it provides insight into the trade-off between speed and accuracy, and opens up opportunities to develop hybrid methods that combine the advantages of both algorithms to create more efficient and effective image processing solutions
Classification of Hepatitis Disease Using The Fuzzy Mamdani Method
Full text linkIn the modern world, almost everyone uses technology and information. This is evident in various fields, ranging from education and employment to entertainment. Society is too dependent on technology and information, to the point of neglecting their own health. There are many diseases caused by neglecting one's own health, one of which is hepatitis. This is because some people pay little attention and are reluctant to get it checked. Because hepatitis is very dangerous for human survival, treatment must begin as soon as the first symptoms appear and assist in the early diagnosis of hepatitis. This will allow for the identification of the type of hepatitis disease. The aim of this research is to apply the Mamdani fuzzy method for the classification of hepatitis diseases. The Mamdani fuzzy method has been successfully utilized in systems for diagnosing hepatitis diseases. In this system, it will provide instructions, namely to select which symptoms are experienced, then you can choose those symptoms by checking them off, and this system will provide a diagnosis based on the symptoms experienced. The diagnosis results include the type of hepatitis disease experienced, as well as treatment solutions. The results obtained for diagnosing hepatitis A disease using fuzzy Mamdani calculation shows that 68% , and the diagnosis of hepatitis B disease using fuzzy mamdani calculations shows 53% , and the diagnosis of hepatitis C disease using fuzzy mamdani calculations shows 59%
Clustering Analysis to Identify Stunting Vulnerability Areas in North Aceh District Using the Fuzzy C-Means Algorithm
Full text linkStunting is a serious public health issue that poses a long-term threat to the quality of human resources. North Aceh Regency is one of the regions with a relatively high prevalence of stunting, requiring targeted and effective intervention strategies. This study aims to classify regions based on their level of stunting vulnerability to support data-driven decision-making. The Fuzzy C-Means (FCM) clustering algorithm was selected due to its ability to handle data with flexible membership degrees, making it suitable for complex classification tasks. The data used in this research were obtained from the North Aceh Health Office for the year 2023 and include variables such as the number of children recorded in the E-PPGBM system, newly entered children in 2023, and the percentages of stunting, wasting, and underweight across 32 subdistricts. The research process involved data collection, literature review, system design and implementation using the Python programming language, and analysis of clustering results. The findings reveal that the 32 subdistricts can be grouped into three main clusters: high vulnerability (13 subdistricts), medium vulnerability (6 subdistricts), and low vulnerability (13 subdistricts). These clusters facilitate the visualization and identification of priority areas requiring more focused stunting interventions. In conclusion, the FCM algorithm proved effective in clustering regions based on stunting-related data. The implication of this study is to provide a foundation for local governments in formulating more efficient and targeted stunting intervention strategies according to the vulnerability level of each area
Implementation and Evaluation of 5G Standalone Network Using Open5GS, srsRAN, and USRP B210 for Research Purposes
Full text linkThis study aims to implement and evaluate an Open Source-based 5G Standalone (SA) network using Open5GS as the Core Network, srsRAN as the Radio Access Network (RAN), and USRP B210 as a Software Defined Radio (SDR) device. A commercial smartphone was used as the User Equipment (UE) to test end-to-end network connectivity and performance. The research method includes software installation, network parameter configuration, system integration, as well as connectivity and performance testing based on ITU-R IMT-2020 standards. The test results show that all network elements were successfully integrated, as indicated by the successful registration and authentication of the UE and the establishment of a data session. Performance testing recorded a downlink throughput of 55 Mbps, uplink throughput of 15 Mbps, latency of 33 ms, jitter of 8.9 ms, and 0% packet loss. Although some performance parameters did not meet the minimum ITU-R IMT-2020 standards, the system proved operable as an independent Open Source and SDR-based solution for experimental purposes in a laboratory environment. Future work should focus on optimizing the backhaul connection, conducting multi-UE testing, and simulating mobility and handover scenarios to assess system performance in large-scale and real-world deployments
Sentiment Analysis to Evaluate Public Service Perception among Surakarta City Residents Using the BiLSTM Model
Full text linkThe growing use of social media as a platform for public communication has opened new opportunities for understanding public opinion regarding government policies, including public services. One of the cities actively discussed on social media is Surakarta, where citizens openly express both appreciation and criticism of local government performance. This study aims to analyze public sentiment toward public services in Surakarta by employing a deep learning-based sentiment analysis approach, specifically using the Bidirectional Long Short-Term Memory (BiLSTM) model. Data were collected from Twitter/X using a web crawling technique with the keywords “pemerintah solo” (Solo government), “kota Surakarta” (Surakarta city), and “kota solo” (Solo city), resulting in 2,168 tweets. The analysis process involved several stages, including preprocessing, sentiment labeling using a lexicon-based method, feature representation with Word2Vec, and classification using five models: SVM, Random Forest, CNN, LSTM, and BiLSTM. The evaluation results show that BiLSTM achieved the best performance with an accuracy of 90.21%, precision of 91.05%, recall of 89.84%, and F1-score of 90.43%. The conclusion of this study is that BiLSTM can effectively classify public sentiment toward public services, especially in the context of informal social media texts. The implication of this research indicates that sentiment analysis can serve as a decision-support tool for designing more responsive and data-driven public policies and provide strategic insights for local governments in improving the quality of public services