TIN: Terapan Informatika Nusantara

Penerapan Semi-Supervised Deep Learning dengan Remixmatch untuk Klasifikasi Penyakit Paru-Paru Menggunakan Citra Chest X-Ray

2026-07-07T23:41:44+07:00

Lung diseases such as pneumonia and COVID-19 viral infection remain significant health problems that require fast and accurate diagnostic processes. The utilization of deep learning-based Computer-Aided Diagnosis (CAD) on Chest X-Ray (CXR) images has demonstrated promising capabilities in assisting disease classification. However, the implementation of deep learning models in the medical field still faces a major challenge, namely the limited availability of labeled data due to the time-consuming annotation process and the involvement of medical experts. This study applies a semi-supervised learning approach using the ReMixMatch algorithm with DenseNet169 architecture as a feature extraction backbone to reduce the dependency on large amounts of labeled data. Experiments were conducted using the public dataset Covid19-Pneumonia-Normal Chest X-Ray Images available on Mendeley Data. The ReMixMatch method utilizes both labeled and unlabeled data through pseudo-labeling, distribution alignment, MixUp augmentation, and consistency regularization mechanisms during the model training process. The evaluation was performed using several labeled data scenarios, namely 10, 20, 30, and 40 labels per class. The experimental results show that the combination of ReMixMatch and DenseNet169 achieved high classification performance with an accuracy of 96.43% on the validation data. The model evaluation obtained a precision value of 96.47%, a recall value of 96.43%, and an F1-score of 96.42%. These results indicate that the semi-supervised learning approach is able to effectively utilize information from unlabeled data, thereby maintaining high Chest X-Ray image classification performance under limited annotation conditions. This study offers an alternative approach to developing a chest X-ray image classification system through the application of the ReMixMatch algorithm combined with the DenseNet169 architecture, enabling the model to achieve good classification performance even with a limited amount of labeled data.

Rancang Bangun Sistem Monitoring Sentimen Berita Media Online Menggunakan IndoBERT Berbasis Web

2026-07-11T19:44:18+07:00

Online news media has become a primary channel shaping public opinion toward government performance, making the ability to monitor news coverage a strategic necessity for institutions such as the Department of Communication and Information (Diskominfo) of West Kalimantan Province. However, the large volume of news coverage renders manual monitoring inefficient and prone to subjectivity. This study aims to design and build a web-based online news sentiment monitoring system named SentimenIQ, which integrates automatic news collection through RSS Feed, sentiment classification using the IndoBERT model, and presentation of analysis results within a single service flow. The system was developed using the Waterfall method with a microservice architecture separating the main Laravel application from the Python FastAPI inference service. Functional testing was conducted using the black box testing method, while classification performance was measured using accuracy, precision, recall, and F1-score derived from a confusion matrix on 120 labeled news articles. The functional testing results show that all system features operated according to requirement specifications, while the classification testing produced an accuracy of 89.17% with a weighted average F1-score of 89.13%. These results prove that the IndoBERT model can be integrated into a web-based operational system and relied upon to monitor news coverage in near real-time, thus serving as a reference for developing similar systems in other government institutions.

Strategi Influencer TikTok untuk Peningkatan Omzet UMKM Berbasis Etika Bisnis Islam

2026-07-11T19:45:34+07:00

Information technology developments have driven social media adoption as a marketing tool for MSMEs. TikTok has proven effective in increasing sales, but its practices need to be examined from Islamic business ethics which prohibit gharar and tadlis and require honesty (shiddiq) and transparency (tabligh). Research integrating TikTok Influencer marketing effectiveness with Islamic business ethics remains limited. This study aims to analyze the effect of TikTok Influencer promotions on MSME Risoles Endul Kraksaan's revenue and review it based on Islamic business ethics. A mixed methods explanatory sequential design was used with 20 respondents (Spearman correlation) and in-depth interviews. Results showed a correlation coefficient of 0.676 (p=0.001), indicating a significant positive relationship between TikTok Influencer promotions and increased revenue. Daily revenue increased from IDR 1-1.7 million to IDR 2.5-3.5 million, equivalent to a monthly increase from IDR 30-51 million to IDR 75-105 million. Promotional practices have fulfilled the principles of shiddiq, tabligh, and are free from gharar and tadlis. The contribution of this research is an integrative model of Influencer marketing effectiveness with Islamic business ethics and an applicable framework for Muslim MSMEs in effective and sharia-compliant digital marketing strategies. This study is limited to one MSME and one platform. TikTok Influencer promotions significantly affect MSME revenue and align with Islamic business ethics.

Pengembangan Aplikasi Berbasis Android untuk Sistem Irigasi pada Sistem Pertanian Cerdas Menggunakan Flutter

2026-07-11T19:46:50+07:00

The agricultural sector plays a crucial role in supporting national food security. However, irrigation management that is still carried out manually often gives rise to various problems, such as water wastage, suboptimal watering, and the high time and labor demands on farmers in monitoring field conditions. Based on field observations, the watering process is still conducted manually, frequently resulting in delayed watering and uncontrolled water usage. The advancement of Internet of Things (IoT) technology provides opportunities to improve irrigation management efficiency through automated monitoring and control systems. This study aims to develop an Android-based smart irrigation system application for agriculture that can assist farmers in monitoring field conditions and controlling the irrigation system remotely. The development method employed is the Prototype model, which is a system development approach consisting of communication, quick planning, design, prototype construction, and testing phases. The application was developed using Flutter and Firebase Realtime Database and integrated with IoT devices to display soil moisture data, control irrigation pumps, display watering history, and provide real-time weather information. System testing was conducted using Black Box Testing to evaluate application functionality and the System Usability Scale (SUS) involving 10 respondents to assess the application's usability level. The Black Box Testing results demonstrated that all application functions operated with a 100% success rate, while the System Usability Scale (SUS) testing yielded an average score of 83, which falls into the Excellent category and Grade A. The findings indicate that the application was successfully developed and is capable of connecting with the IoT system in real-time, thereby facilitating more effective and efficient land monitoring and irrigation management. The contribution of this research is the development of an Android-based smart irrigation system application that integrates soil moisture monitoring, automatic and manual pump control, watering scheduling, notifications, watering history, and real-time weather information within a single platform connected to the Internet of Things (IoT).

Pengembangan dan Evaluasi Usabilitas Gim Edukasi Bahasa Jawa Menggunakan Game Development Life Cycle

2026-07-11T19:48:00+07:00

Javanese is a regional language that should be introduced from elementary school because it represents cultural values, local identity, and Javanese manners. This study aims to design, develop, and evaluate the usability of an Android-based Javanese learning game using the Game Development Life Cycle (GDLC). The study was conducted at SDN Candirejo by involving the school principal as an interview informant and 15 students from grades III, IV, and V as respondents. The game content was prepared with reference to the Remen Basa Jawi 2013 Curriculum textbooks for grades I to VI and input from the school. The game applies the Drill and Practice method in an adventure puzzle format through area exploration, vocabulary collection, practice questions, a dictionary, inventory, and boss-stage evaluation. The content covers daily vocabulary, the use of Ngoko and Krama, identification of mixed-language sentences, and language selection based on the interlocutor. Alpha testing using Black Box Testing showed that all 16 scenarios worked as expected, with a success rate of 100%. Beta testing using the System Usability Scale (SUS) produced an average score of 72.17, which falls into the Acceptable category with a Good rating. These findings indicate that the game has good usability and can be considered an alternative medium for supporting Javanese language practice in elementary schools. This study contributes an interactive digital learning medium that integrates vocabulary practice, Ngoko-Krama use, and contextual language selection to support Javanese language preservation in elementary schools.

Pengembangan Gim Edukasi Pengenalan Kosakata Multibahasa Berbasis Web untuk Anak Usia Dini Menggunakan Metode Multimedia Development Life Cycle (MDLC)

2026-07-11T19:43:19+07:00

Language development is a fundamental aspect of early childhood growth, particularly for children aged 4–6 years, a period widely recognized as the golden age. However, multilingual vocabulary learning in Early Childhood Education (ECE) institutions still faces several challenges, including the predominance of conventional learning methods, limited interactivity, and insufficient ability to maintain children's motivation and engagement throughout the learning process. As a result, vocabulary acquisition across multiple languages has not been optimally achieved. This study aims to develop a web-based multilingual vocabulary learning game for early childhood using the Multimedia Development Life Cycle (MDLC) method, which consists of six stages: Concept, Design, Material Collecting, Assembly, Testing, and Distribution. The game was developed using the Next.js framework with TypeScript and features five game modes: Matching Pairs, Treasure Box, Quiz Time, Word Builder, and Audio Match, supporting three languages: Indonesian, English, and Arabic. The study was conducted at PG-TK Islam Al-Azhar Cairo Yogyakarta, involving 14 six-year-old children as observation participants and two teachers as respondents for the System Usability Scale (SUS) evaluation. Functional testing using Black Box Testing across 21 test scenarios showed that all system functions operated successfully. The usability evaluation resulted in an average SUS score of 81.25, which falls into the Good and Acceptable categories. Furthermore, observations indicated an average game usage achievement of 93.57% and an average vocabulary recognition achievement of 92.86%, both classified as Very Good. These findings demonstrate that the developed game is feasible as a multilingual vocabulary learning medium for early childhood education in terms of functionality, usability, children's engagement, and effectiveness in supporting vocabulary acquisition.

Pengembangan Sistem Arsip Dinamis Kontrak Tower Menggunakan Metode DevOps Berbasis Web

2026-07-12T16:46:30+07:00

The management of tower lease renewal warning letters at PT Indosat Ooredoo Hutchison (IOH) faces various challenges, including process inefficiency, recording errors, and limited data visibility due to reliance on manual processes that have not been systematically structured. These conditions have led to slow administrative responses and an increased risk of delays in handling alternative tower replacement mitigation. This study aims to design and develop a web-based dynamic archive information system using the DevOps method to improve software development effectiveness while optimizing operational services within the Tower Lease Provider (TLP) department at PT IOH. DevOps was selected for its ability to continuously integrate development and operations processes, thereby accelerating software release cycles and improving system quality on an ongoing basis. The research was conducted at PT IOH's Jakarta office during the period of April–June 2026, involving four tower providers as informants and one internal TLP representative. Research stages included requirements analysis, database design using Entity Relationship Diagrams (ERD) and Physical Data Models (PDM), implementation of monitoring features, automated notifications, and reporting, through to deployment via the TLP SYSTEM ARSIP application. Testing was carried out using the Black Box Testing method, confirming that all core functions performed as expected. The results demonstrate that the system operates stably, with improved recording accuracy, a reduction in administrative processing time of up to 60–70%, and a decrease in recording errors of up to 80% compared to the previous manual process. Through DevOps implementation, PT IOH has acquired a reliable and adaptive digital platform responsive to evolving operational needs in tower lease renewal management.

Perbandingan Metode Random Forest dengan Decision Tree pada Sistem Rekomendasi Olahraga Berdasarkan Karakteristik Kepribadian

2026-07-12T16:48:01+07:00

Selecting sports that match individual characteristics is an important factor in increasing motivation, comfort, and consistency in performing physical activities. Each individual has different psychological characteristics, activity preferences, exercise intensity levels, and desired exercise goals. One approach that can be used to provide more personalized sports recommendations is by utilizing personality characteristics based on the Big Five Personality (OCEAN) model. This study aims to compare the performance of the Decision Tree and Random Forest algorithms in classifying sports categories based on personality characteristics, exercise intensity, social preferences, exercise location, and exercise goals. The novelty of this research lies in the implementation and comparative evaluation of these two algorithms in a Big Five Personality-based sports recommendation system, which has not been widely developed. The dataset used was obtained from sports experts, consisting of 603 records containing personality attributes and supporting factors related to sports activities. The dataset was divided using the train-test split method with a proportion of 67% training data and 33% testing data. The research stages included data validation, categorical attribute transformation using Ordinal Encoding, classification model development, and evaluation using accuracy, precision, recall, F1-score, and confusion matrix metrics. The results showed that the Random Forest algorithm achieved better performance than Decision Tree, with an accuracy of 82.91%, precision of 0.89, recall of 0.83, and F1-score of 0.81. Meanwhile, Decision Tree obtained an accuracy of 77.89%, precision of 0.65, recall of 0.78, and F1-score of 0.70. These results indicate that the ensemble approach in Random Forest is capable of capturing more complex data patterns and producing more accurate sports category classifications. This research is expected to serve as a foundation for developing a more adaptive and personalized sports recommendation system based on user characteristics.

Pengembangan Sistem Informasi Manajemen Data Gapoktan Menggunakan Metode Fountain

2026-07-16T01:27:51+07:00

Data management in Farmers Group Associations (Gapoktan) is still largely done manually, causing various problems, such as data duplication, difficulty in finding information, delays in preparing reports, and a high risk of document loss. This condition also exists in Gapoktan in the Sokaraja area, Banyumas Regency, which requires an integrated system to support more effective and efficient organizational data management. This study aims to develop a web-based Gapoktan Data Management Information System using the Fountain method as a software development model. The study used a Research and Development (R&D) approach with the research location at Gapoktan in the Sokaraja area, Banyumas. Data collection was carried out through observation, interviews, and documentation studies. The research respondents numbered 18 people selected using a purposive sampling technique, consisting of the chairperson, secretary, and treasurer of Gapoktan. System testing was conducted using Black Box Testing to test system functionality, while user evaluation was conducted through questionnaires to measure ease of use, system effectiveness, work efficiency, and user satisfaction. The results showed that all system functions ran as needed with a test success rate of 100%. The user evaluation results obtained an average score of 90.00%, categorized as very good, consisting of aspects of ease of use of 91.11%, system effectiveness of 88.89%, work efficiency of 86.67%, and user satisfaction of 93.33%. These results indicate that the developed system is able to improve the quality of data management, accelerate the administrative process, and support more effective report preparation within the Gapoktan environment.

Analisis Kinerja Protokol ESP-NOW Sebagai Komunikasi Data Multi-Node Wireless Sensor Network Pada Esp32 dengan IoT Gateway dan Visualisasi Web

2026-07-12T16:50:54+07:00

IoT-based Wireless Sensor Networks (WSN) are typically low-rate and latency-sensitive, making the selection of an efficient communication protocol critical. Unlike prior studies that generally examined a single aspect of ESP-NOW performance in isolation (range, comparison with Wi-Fi, or delay in multi-node scenarios), this study contributes a novel integrated evaluation framework that simultaneously analyzes ESP-NOW performance on ESP32 with a many-to-one topology across six operational variables: number of nodes (1–2), transmission interval (1, 3, and 5 seconds), transmission distance (10–50 meters), environmental condition (with and without obstacles), device positioning, and data load, while integrating the system end-to-end with an ESP32-based IoT gateway, InfluxDB time-series database, and real-time Grafana visualization. Each sensor node is equipped with a GY-BMP280 sensor for temperature and air pressure measurement. Performance was evaluated using TIPHON-standard Quality of Service (QoS) parameters, including delay, throughput, jitter, and packet loss, along with additional RSSI and power consumption measurements. Results show that a 1-second transmission interval yields the best QoS with delay <150 ms, jitter 0–75 ms, and 0% packet loss at distances up to 30 meters without obstacles. The IoT gateway successfully received data without packet loss under low-to-medium loads and demonstrated a write capacity to InfluxDB of up to 1000 data points per second. The main contribution of this study is the provision of a TIPHON-standardized ESP-NOW performance map and an end-to-end system architecture that can serve as a configuration reference for small-to-medium-scale IoT developers. This study is limited to a maximum of two nodes, so generalization to larger scales requires further investigation.

Pengembangan Multimedia Interaktif Berbasis Visual Tiga Dimensi untuk Pembelajaran Numerasi Anak Tunagrahita

2026-07-12T16:52:18+07:00

This study aims to develop three-dimensional visual-based interactive multimedia to be used as a learning aid for numeracy among children with intellectual disabilities. The study was conducted at the Cahaya Impian Masa Depan (CIMD) Foundation. The research method employed the Multimedia Development Life Cycle (MDLC), which consists of six stages: concept, design, material collecting, assembly, testing, and distribution. The developed multimedia is capable of displaying three-dimensional animal visualizations, interactive audio, numeracy quizzes, adn a reward system,which are designed to help students understand numerical concept more concretely. Testing was conducted through three evaluations: a content expert review, a media expert review, and a user responses evaluation using the Usability Metric for User Experince (UMUX) method. The study participants consisted of two teachers and 15 students with intellectual disabilities as users of the learning media. The content expert review yielded a validity percentage of 92,86%, classified as highly valid, and the media expert review yielded a validity percentage of 90%, classified as highly valid. Additionally, the user response evaluation yielded a UMUX score of 84,16% which falls into the Excellent category and Grade A. The research findings indicate that three-dimensional visual-based interactive multimedia possesses excellent usability, is easy to use, and is suitable as an alternative learning medium for numeracy education for children with intellectual disabilities. This study contributes by providing a learning medium designed based on the characteristics of children with intellectual disabilities and the learning needs identified at the Cahaya Impian Masa Depan (CIMD). The developed multimedia is expected to serve as a reference for the development of interactive learning media in inclusive education as well as special education institutions.

Implementasi Tata Kelola Teknologi Informasi dalam Transformasi Digital Manajemen Sumber Daya Manusia: Tinjauan Literatur

2026-07-16T15:12:49+07:00

In the era of rapidly accelerating digital transformation, organizational success in adopting technology for human resource management (HRM) depends not only on technological infrastructure but also on effective IT governance that bridges strategy and implementation. This study addresses the persistent execution gap between strategic HRIS/MIS/ICT adoption policies and their implementation by HR staff, caused by limited digital competencies, weak operational guidelines, and inadequate adaptive governance. It aims to identify factors influencing IT implementation effectiveness in HRM, digital competency development strategies integrated with IT governance, and governance frameworks supporting digital HRM transformation. A Systematic Literature Review (SLR) following the PRISMA 2020 protocol was conducted. From 865 articles retrieved from the ScienceDirect database, 41 studies met the eligibility criteria and were analyzed. The findings indicate that IT implementation effectiveness is shaped by four key factors: digital transformational leadership, infrastructure and data governance readiness, adaptive organizational culture, and technical-ethical risk management. Effective digital competency development combines structured reskilling/upskilling programs, AI-enabled adaptive learning, and the enhancement of AI literacy alongside uniquely human skills. Furthermore, successful digital HRM transformation requires integrating ambidextrous governance, ethical data governance, and maturity model-based roadmaps. This study proposes the IT Governance–HRM Digital Transformation Integration Framework (IGHDTIF) as a holistic conceptual model integrating these dimensions and addressing a gap identified in the reviewed literature.

Deteksi Port Scanning pada Jaringan IoT dengan Pemantauan Web Real-Time

2026-07-16T15:36:28+07:00

Port scanning is a reconnaissance activity that may precede service exploitation on Internet of Things (IoT) devices. This study designs and implements a prototype for detecting TCP port scanning with real-time web monitoring in the local network of PT WIMISEC, Bekasi. The system was developed using a Research and Development approach and Waterfall stages, integrating a Laravel application, a MySQL database, and a Python detection engine. Detection applies a threshold rule based on the number of unique destination ports accessed by one source IP address within a time window. The evaluation covered functional testing, five TCP SYN scans, five TCP Connect scans, three browsing activities, and three normal SSH logins. All 10 scan trials generated alerts (an observed detection rate of 100%), while none of the six normal trials generated an alert (an observed false-positive rate of 0% in the limited sample). The mean latency from detection-engine identification to dashboard alert display was 1.24 seconds. These results demonstrate the prototype's functional feasibility for initial LAN monitoring rather than general accuracy across attack patterns. The evaluation remains limited by its small sample, the absence of mixed-traffic and stress testing, and the lack of tests involving UDP, slow, distributed, and evasive scans.

Analisis Komparatif Konfigurasi Multilayer Perceptron pada Classifier Head RoBERTa untuk Klasifikasi Ujaran Kebencian

2026-07-20T00:12:23+07:00

The widespread dissemination of hate speech and offensive language on social media has increased the demand for accurate automated text classification systems. Although RoBERTaForSequenceClassification has been widely used for various for text classification task, the effect of its default classifier head configuration on classification performance has not yet been systematically evaluated. As the main contribution, this study conducts a controlled evaluation of 32 multilayer perceptron (MLP)-based classifier head configurations, varying the number of hidden layers, activation functions, and dropout rates, against the default classifier head on the English HASOC 2021 dataset for two subtasks: binary and multiclass classification. Each configuration was evaluated using Stratified 5-Fold Cross-Validation with Macro-F1 as the evaluation metric, after which the best-performing configuration was further evaluated on an independent test set. For the binary task, the best configuration achieved a test Macro-F1 of 80.90%, about 0.3 percentage points higher than the baseline's 80.59%. For the multiclass task, the configuration with the highest validation performance instead achieved a test Macro-F1 of 65.68%, about 0.4 percentage points lower than the baseline's 66.11%, showing that an advantage observed during cross-validation does not always hold on the test set. Further analysis revealed that excessively deep hidden layers combined with aggressive dimensional compression can sharply degrade performance on the multiclass task. These findings indicate that the effect of classifier head configuration is small and task-dependent, so systematic evaluation remains necessary before adopting a given configuration in place of the default classifier head when fine-tuning RoBERTa-based models.

Analisis Pengaruh Peningkatan Kualitas Citra Berbasis Histogram terhadap Kinerja ResNet-18 pada Klasifikasi Karakter Katakana

2026-07-20T00:13:53+07:00

Handwritten Katakana character classification remains a challenging task due to variations in individual writing styles and suboptimal image quality, such as low contrast and unclear character details. These conditions can affect the model's ability to extract discriminative features and accurately recognize characters. This study aims to analyze the impact of histogram-based image enhancement techniques, namely Histogram Equalization (HE), Adaptive Histogram Equalization (AHE), and Contrast Limited Adaptive Histogram Equalization (CLAHE), on the performance of the ResNet-18 model for Katakana character classification. The ETL5 dataset, consisting of 9,200 handwritten character images across 46 character classes, was used in this study. The dataset was divided into training and testing sets with an 80:20 ratio. To increase data diversity and reduce the risk of overfitting, data augmentation using rotation and translation was applied to the training set. The classification process employed a transfer learning-based ResNet-18 model initialized with ImageNet pretrained weights, where only the fully connected layer was fine-tuned using the Adam optimizer for 50 epochs. The experimental results show that CLAHE achieved the best performance with an accuracy of 97.01%, while HE and AHE obtained accuracies of 83.15% and 76.79%, respectively. These findings indicate that CLAHE is more effective than HE and AHE in improving the classification performance of handwritten Katakana characters using ResNet-18, resulting in more accurate predictions. This study contributes by providing a comparative analysis of the effects of three histogram-based image enhancement methods on the performance of ResNet-18 for handwritten Katakana character classification.

Pengembangan Sistem Informasi Penanganan Feedback Client Berbasis Web Menggunakan Metode Waterfall

2026-07-20T00:15:14+07:00

Client feedback management at BIIS Corp still relies on WhatsApp group communication and manual spreadsheet recording, resulting in difficulties in monitoring work progress, inadequate documentation of handling history, and poor coordination between staff and programmers. This study aims to design and develop a web-based client feedback information system to support centralized and structured feedback management. Problem identification was conducted using the PIECES framework, which revealed weaknesses across all six aspects. The system was developed using the Waterfall model within the SDLC framework, with UML-based modeling and functional testing through the Blackbox Testing method. Built using PHP, MySQL, HTML, CSS, and JavaScript, the system accommodates three user roles, namely administrator, staff, and programmer, and is equipped with features including feedback recording, programmer assignment, status updates, completion confirmation, reopen feedback mechanism, real-time notifications, activity log, and performance reporting. Testing across 15 scenarios achieved a 100% success rate. The contribution of this research is the development of a system specifically designed for a software house environment, integrating three user roles into a centralized workflow equipped with a reopen feedback mechanism, real-time notifications, and an activity log to support more effective feedback monitoring and documentation.

Sistem Monitoring Energi Genset pada Stasiun Penyiaran TVRI Berbasis IoT

2026-07-20T00:16:40+07:00

The availability of electrical energy is a crucial factor in maintaining the continuity of broadcasting station operations. Public Broadcasting Institution (LPP) TVRI North Sulawesi, where a generator set (genset) is used as a backup power source during interruptions to the main electrical supply. The primary challenge is the limited capability of conventional monitoring systems to observe electrical parameters, including current, power, energy consumption, and power source status, in real time. This study aims to develop an Internet of Things (IoT)-based generator energy monitoring system using an ESP32 microcontroller, SCT-013 current sensor, Liquid Crystal Display (LCD), and cloud-based monitoring platforms. The proposed system measures electrical parameters in real time and transmits the data via a Wi-Fi network to the Blynk application and Google Spreadsheet, enabling remote monitoring by operators. Experimental results show that the proposed system measures electrical current with a measurement error ranging from 1% to 8%, yielding an average error of 4.17% and an average measurement accuracy of 95.83% compared with a reference clamp meter. The system also successfully calculates electrical power up to 2,640 W at a current of 15 A, automatically records monitoring data in Google Spreadsheet, and displays real-time information through the Blynk application when a Wi-Fi connection is available. Furthermore, the system accurately detects PLN ON, Generator ON, and both sources OFF conditions based on current variations in the Automatic Transfer Switch (ATS). The main contribution of this study is the development of an integrated IoT-based generator energy monitoring system capable of measuring electrical current, calculating power and energy consumption, detecting PLN–generator switching status through the ATS, storing historical data in Google Spreadsheet, and providing real-time monitoring and notifications via the Blynk platform to support reliable energy management in broadcasting stations. These results demonstrate that the proposed system provides reliable real-time generator monitoring and improves the effectiveness of monitoring, data logging, and maintenance activities, thereby enhancing the reliability of the electrical system at TVRI North Sulawesi Broadcasting Station.

Analisis Probabilitas Aroma Sugar browning Kopi Robusta pada Proses Roasting Tradisional Menggunakan Naive Bayes

2026-07-20T00:18:34+07:00

The roasting process is one of the key stages in coffee processing because it greatly influences aroma and final flavor development. In Pasemah Air Keruh District, Robusta coffee is still roasted manually, and no clear guidelines exist for roasting conditions that consistently produce a sugar browning aroma. This study aimed to analyze the probability of sugar browning aroma formation in Robusta coffee and identify the dominant roasting category using the Naive Bayes algorithm. Data were collected from 150 Robusta coffee processors in Pasemah Air Keruh District. The observed variables included heat intensity, roasting duration, stirring, and cooling. Aroma classes were determined by baristas as the ground truth and categorized into Medium Roast, Medium Dark Roast, and Dark Roast. The data were cleaned, validated, converted into numerical form, and divided into 120 training and 30 testing samples. The results showed that the Naive Bayes algorithm achieved an accuracy of 76.67% and identified the relationship between roasting variables and sugar browning aroma. The Medium Dark Roast category had the highest probability of producing sugar browning aroma. This study provides an overview of the relationship between traditional roasting parameters and the probability of sugar browning aroma formation, which may serve as a reference for developing a more consistent coffee roasting process.

Perbandingan Algoritma K-Means dan Fuzzy C-Means pada Segmentasi Citra Biji Jengkol Deteksi Kematangan

2026-07-20T00:20:07+07:00

This study aims to compare the performance of K-Means and Fuzzy C-Means (FCM) algorithms in image segmentation of jengkol seeds (Archidendron pauciflorum) for automatic ripeness detection. The dataset comprises 300 images categorized into three ripeness classes: ripe (100 images), half-ripe (100 images), and unripe (100 images). Images were acquired using a 12 MP smartphone camera at a standardized resolution of 640×480 pixels under controlled lighting at a distance of 20 cm from the object. The research pipeline includes image preprocessing (RGB-to-HSV and LAB/CIELAB color space conversion, median filter noise reduction, and contrast enhancement), K-Means and FCM segmentation, color and texture feature extraction using the Gray Level Co-occurrence Matrix (GLCM), and performance evaluation based on accuracy, Peak Signal-to-Noise Ratio (PSNR), and computational time. Results indicate that FCM achieves 90–93% accuracy and 30–32 dB PSNR, outperforming K-Means (85–88% accuracy, 27–29 dB PSNR). Nevertheless, K-Means excels in computational efficiency (0.45 s vs. 1.20 s). FCM is recommended for high-accuracy applications, whereas K-Means is preferred when computational efficiency is prioritized.

Rancang Bangun Sistem Informasi Monitoring, Evaluasi, dan Pembinaan Tata Kelola Sekolah Dasar Menggunakan Metode Waterfall

2026-07-20T00:24:57+07:00

The implementation of monitoring, evaluation, and guidance for elementary school governance in the Regional Education Office still faces challenges in information management because data from schools has not yet been integrated, resulting in evaluation, guidance, and decision-making processes that are not yet effective. This study aims to design and develop a web-based information system to support the integrated management of monitoring, evaluation, and guidance using the Waterfall method, which includes needs analysis, design, implementation, testing, and conceptual maintenance. The research resulted in a system capable of linking the monitoring, evaluation, and guidance processes into a single operational mechanism for information management, so that monitoring results directly serve as the basis for conducting evaluations, while evaluation results serve as the basis for developing guidance without requiring reprocessing or transferring data to other media. Unlike similar systems, which generally still separate these three processes, the developed system forms a continuous information management cycle, thereby supporting data consistency and strengthening the Regional Office’s decision-making in overseeing elementary school governance. The results of Black Box Testing showed that all system functions operate in accordance with functional requirements, while User Acceptance Testing (UAT) achieved a user acceptance rate of 83%, indicating that the system is acceptable and supports users’ operational needs. Thus, this research not only produced a web-based information system but also offered an operational mechanism that integrates monitoring, evaluation, and guidance into a single information management cycle to support more effective oversight of elementary school governance.

Pengembangan FMEWS Berbasis IoT dan MQTT untuk Peringatan Banjir

2026-07-20T00:29:09+07:00

Floods are disasters that frequently cause physical, social, and economic losses, including in Buleleng Regency, Bali. The main problem in flood mitigation lies in delayed early warning delivery and the absence of a centralized platform that integrates field monitoring data with operational disaster management data. This study aims to design and develop the Flood Monitoring and Early Warning System (FMEWS) based on the Internet of Things (IoT) and the Message Queuing Telemetry Transport (MQTT) protocol to support real-time flood monitoring and early warning. The system was developed using the Waterfall model of the System Development Life Cycle (SDLC) and consists of IoT devices, an IoT Gateway, and a web-based information system integrating water level monitoring, early warning notifications, shelter management, public facility management, community reports, and spatial visualization. The novelty of this study lies in integrating real-time IoT-based flood monitoring with operational disaster management data into a single centralized platform. The results show that a transmission rate of 2–5 Hz provides the best balance between latency and communication reliability, with a Packet Delivery Ratio (PDR) of at least 96.67%. Usability testing involving 22 respondents produced an average System Usability Scale (SUS) score of 70.0, which falls into the Acceptable category. These findings indicate that FMEWS is feasible as a supporting platform for flood monitoring and early warning in Buleleng Regency.

Pengembangan Kalkulator HPP Penentu Harga Jual Berbasis Generative AI Menggunakan Agile dan Cost-Plus Pricing

2026-07-20T00:34:59+07:00

Determining the cost of production is an important aspect of business because it directly affects the setting of product selling prices. Errors in production cost calculation often arise due to users’ limited understanding, inconsistent cost recording, and manual calculations that are prone to mistakes. The implementation of an information system provides an effective solution to reduce these risks. This study aims to design and develop a web-based production cost calculator application (HPP) that enables users to calculate production costs and determine product selling prices using the Cost-Plus Pricing method. The system development follows the Agile model, which includes planning, design, development, testing, and evaluation phases. The use of Artificial Intelligence (AI) in the development phase enables the creation of a more efficient and economical digital system with almost no additional cost. Generative AI is utilized during the development phase to assist in preparing the product backlog, creating the initial program code design, and accelerating the process of identifying and fixing errors through debugging. The results show that the developed HPP application can be used directly by the public without requiring software installation or additional database configuration, and can be accessed free of charge through the provided link. The contribution of this study is to provide a practical, economical, and easily accessible HPP calculator application that can be used by various types of businesses as a tool for determining selling prices based on production costs and target profit, while also demonstrating the use of Generative AI in accelerating the application development process efficiently. The black-box testing results show a success rate of 100%, indicating that all main application functions operate according to the specified requirements.

Evaluasi Tata Kelola TI pada Industri Manufaktur Kereta Api dalam Proses Pengadaan Menggunakan COBIT 2019

2026-07-20T13:54:06+07:00

The procurement process in the railway manufacturing industry exhibits high complexity due to product characteristics that vary in specification for every project. PT Industri Kereta Api (INKA), a strategic state-owned enterprise (SOE) located in Madiun, East Java, faces challenges in information technology (IT) governance to support this process, while simultaneously being bound by regulatory mandates for independent IT maturity assessments. This study aims to measure the maturity level of IT governance in PT INKA's procurement process and formulate measurable improvement recommendations. The evaluation was conducted using the COBIT 2019 framework across five strategic domains: APO10 (Managed Vendors), APO12 (Managed Risk), BAI03 (Managed Solutions Identification and Build), BAI06 (Managed IT Changes), and MEA01 (Performance and Conformance Monitoring). Utilizing a descriptive approach, this research involved purposive stakeholders from PT INKA's IT department. Data collection was carried out through semi-structured interviews, field observations, document analysis, and the COBIT 2019 assessment instrument based on the NPLF scale. The evaluation results reveal that all domains are currently at Level 3 (Established Process), indicating a 1-level gap from the targeted Level 4 (Predictable Process). The primary obstacle is the lack of consistent quantitative measurement implementation, such as standardized vendor KPI (Key Performance Indicators ), threshold-based KRI (Key Risk Indicators), and an integrated business intelligence platform. The resulting recommendations are practical and phased, with priority given to the establishment of a Change Advisory Board. This study contributes empirical evidence on IT governance evaluation specifically addressing procurement in the railway manufacturing industry, characterized by high product customization, and serves as a practical reference for similar industry in developing capability improvement measures toward a more predictable process level.

Sistem Pakar Berbasis Certainty Factor untuk Identifikasi Tingkat Stres Akademik dan Rekomendasi Konseling Mahasiswa

2026-07-22T00:11:55+07:00

Academic stress is a common problem experienced by students due to high academic demands, which impacts academic achievement and mental health. Therefore, an early detection mechanism is needed to help identify the level of academic stress quickly and objectively as a basis for providing appropriate counseling services. This study aims to develop a Certainty Factor-Based Expert System to identify students' academic stress levels and provide counseling recommendations based on the diagnosis results. The Certainty Factor method is implemented by combining the Expert Certainty Factor (Expert CF), User Certainty Factor (User CF), and Rule Certainty Factor (Rule CF) to accommodate uncertainty in the inference process, while the knowledge base is compiled through knowledge acquisition from guidance and counseling experts based on the Perception of Academic Stress (PAS) indicator. The results show that the system is able to identify academic stress levels based on the symptoms selected by students, generate a Certainty Factor value as a level of confidence in the diagnosis, and provide counseling recommendations that are appropriate to the identified stress categories. In the test scenario, the system generated a Certainty Factor value of 77.6%, indicating a Moderate Stress category (S2). This research contributes to the development of an expert system that integrates the identification of academic stress levels and counseling recommendations in one web-based application so that it can support the early detection process and decision-making for students, academic advisors, and counselors.

Analisis Performa dan Efisiensi VGG16, ResNet50, dan MobileNetV2 pada Klasifikasi Citra Jajanan Tradisional

2026-07-22T01:38:56+07:00

Previous research on traditional food image classification demonstrated that the VGG16 architecture could achieve good classification performance. However, the study focused on a single architecture and did not provide a comprehensive comparison between predictive performance and computational efficiency across multiple models. This limitation highlights the need to evaluate other architectures in order to identify a model that is not only accurate but also suitable for different deployment environments. This study aims to compare the performance and efficiency of three transfer learning architectures, namely VGG16, ResNet50, and MobileNetV2, for traditional food image classification. The dataset consisted of 2,445 images grouped into ten classes: batun bedil, bubur injin, jaje lukis, jaje piling, jaje wajik, kaliadrem, klepon, laklak, ongol-ongol, and pisang rai. The dataset was divided using a 70:15:15 ratio into 1,706 training images, 363 validation images, and 376 testing images. All models were trained under the same configuration, using an input size of 224 × 224 pixels, batch size of 64, Adam optimizer, learning rate of 0.0001, categorical crossentropy loss, data augmentation, and 10 epochs. Model evaluation was conducted using accuracy, precision, recall, F1-score, training time, model size, and inference time per image. The results showed that ResNet50 achieved the best classification performance, with an accuracy of 84.04%, precision of 85.15%, recall of 85.16%, and F1-score of 84.89%. MobileNetV2 achieved an accuracy of 82.45% with the smallest model size of 12.98 MB, while VGG16 obtained an accuracy of 73.67%. These findings indicate that ResNet50 is more suitable for systems that prioritize classification performance, whereas MobileNetV2 is more appropriate for applications requiring a lightweight model. The contribution of this study lies in the comparative evaluation of performance and efficiency across three architectures under the same experimental setting, resulting in practical model recommendations based on deployment requirements.

Pengembangan Sistem Ticketing Berbasis Web Menggunakan Metode Waterfall untuk Manajemen Layanan TI

2026-07-28T06:49:59+07:00

Systematic service management is crucial for internet service providers. PT Ikhlas Cipta Teknologi faces the problem of decentralized management of customer complaints and service requests due to its reliance solely on the WhatsApp instant messaging app. This practice often leads to missed reports, loss of problem resolution records, and unmeasurable technician handling times. The objective of this research is to design and build a web-based internal Ticketing Information System tailored to the company's operational needs to address these issues. The system was developed using the Waterfall method, utilizing the Next.js framework, and the MongoDB Atlas database. The proposed solution includes digitizing complaints into structured tickets, a deadline monitoring mechanism through a Service Level Agreement (SLA), and a four-level escalation flow: Customer Service, Helpdesk, Network Operation Center (NOC), and Super NOC. Black Box Testing evaluation results demonstrated that all system functionality ran smoothly. Furthermore, the results of the feasibility test from end users through User Acceptance Testing (UAT) recorded a percentage of 93.6%. This figure indicates a very good level of system acceptance. The implementation of this system is able to improve the management of complaints or customer service requests by converting reports into structured data, providing a more transparent handling history, and monitoring resolution times accurately, thereby supporting the improvement of the company's operational effectiveness. The research contributes by integrating Service Level Agreements (SLAs), Role-Based Access Control (RBAC), and a four-tier escalation mechanism into a web-based internal ticketing system, facilitating more structured incident management within an ISP environment.

Pengembangan Automotive-Hub Platform Ekosistem Otomotif Menggunakan Metode Agile

2026-07-28T06:51:17+07:00

Current automotive business processes in Batam, encompassing both sales and rentals, remain dominated by conventional methods and the fragmented dissemination of information across social media platforms. This reliance leads to search time inefficiencies, restricted consumer access to specific information, and a distinct lack of an integrated information system capable of accommodating all entities within the automotive business ecosystem. This study aims to bridge this gap by developing an Integrated Automotive-Hub Platform designed to seamlessly and inclusively connect the automotive ecosystem—specifically, industry entrepreneurs and consumers. The software development lifecycle implements the Agile methodology through iterative cycles (Sprints), comprising product backlog creation, iterative development, and continuous testing. Built on a Model-View-Controller (MVC) architecture, the platform features robust core functionalities, including Create, Read, Update, and Delete (CRUD) product management, Peer-to-Peer (P2P) Live Messaging, a Wishlist system, and a Rating & Review module. System evaluation was conducted using Black-Box Testing to validate software functionality, alongside comparative operational time testing to measure the productivity of the new system against traditional manual processes (which utilize messaging applications and spreadsheets). Functional testing results indicate that 14 system scenarios executed perfectly, with no interface or database errors detected. Furthermore, comparative time analyses demonstrate that the platform drastically reduces operational durations within the search and fleet management modules, achieving a Task Success Rate of 95% to 100%. The primary contribution of this research is the engineering of a centralized, public-facing digital automotive ecosystem. This platform tangibly optimizes information transparency (inclusive information), accelerates transaction processes, and minimizes human error within Batam's automotive industry.

Perancangan dan Implementasi Aplikasi Mobile Buku Kas Menggunakan Metode User-Centered Design untuk Mendukung Pengelolaan Keuangan Para Pelaku UMKM

2026-07-28T06:53:07+07:00

The rapid development of information technology has provided convenience in various aspects of life, including financial management. Proper financial management is one of the determining factors for business sustainability and growth, particularly for Micro, Small, and Medium Enterprises (MSMEs), which serve as one of the main pillars of the national economy. However, many MSME owners still do not have the habit of recording daily transactions regularly due to limited time and the lack of simple recording tools. This problem affects the ability of MSME owners to accurately monitor cash flow, distinguish personal and business finances, and make business decisions based on valid data. Manual recording using physical cash books is also considered inefficient, prone to human error, easily lost or damaged, and difficult to recapitulate or analyze quickly when needed. This condition is further exacerbated by the low level of digital literacy among some MSME owners and the limited availability of financial recording applications that are truly suitable for the characteristics of micro-enterprises, namely simple, lightweight, easy to learn, and accessible anytime through mobile devices without requiring special training. This study aims to design and develop a mobile cash book application to assist MSME owners in recording daily income and expenses efficiently. The method used is User-Centered Design (UCD), which places user needs as the main focus throughout the process, including requirements identification, interface design, prototyping, and user evaluation. The application was developed using Kotlin in Android Studio with local data storage using SQLite synchronized with a MySQL database on the server side. Functional testing was conducted using the Black Box Testing method, while user acceptance evaluation was performed through a Likert-scale questionnaire involving 35 respondents based on three aspects: Usability, UI/UX, and Efficiency. The results showed that the application functioned properly and received very positive responses from users across all evaluation aspects, demonstrating that the implementation of UCD can improve user satisfaction in independently managing finances through mobile devices. This research contributes to the development of an Android-based mobile cash book application that provides features for recording income and expenses, transaction category management, financial report visualization, PDF report export, and transaction recording reminders. Furthermore, the implementation of synchronization mechanisms between SQLite and MySQL enables transaction recording even when an internet connection is unavailable, with data updates performed when the connection becomes available again.

Implementasi Multimodal Emotion Recognition Menggunakan CNN-BiLSTM-Attention pada Audio dan Mobilenetv2 pada Ekspresi Wajah

2026-07-28T06:55:28+07:00

The advancement of artificial intelligence technology has significantly contributed to the development of emotion recognition systems capable of understanding human emotional states more accurately. However, unimodal approaches that rely on a single data source, such as speech or facial expressions, still face limitations when dealing with dynamic environmental conditions. This study aims to implement a Multimodal Emotion Recognition system by integrating audio and visual modalities to improve emotion detection accuracy. The audio modality is processed using a CNN-BiLSTM-Attention architecture with Mel-Spectrogram representations as input, while the visual modality is analyzed using MobileNetV2 to recognize facial expressions. The integration of both modalities is performed using the Late Fusion method at the decision level through a weighted fusion approach, assigning weights of 0.4 and 0.6 to the audio and visual modalities, respectively. This research employs an experimental method with a quantitative approach. The research stages include literature review, data collection, audio and video preprocessing, model design, model training, multimodal integration, and system evaluation. Audio data are transformed into Mel-Spectrogram features, while video data are processed using MediaPipe Face Detection to extract facial regions prior to emotion classification. The proposed system is designed to recognize three emotional classes: neutral, happy, and angry. Model performance is evaluated using a Confusion Matrix, accuracy, precision, recall, and F1-score metrics. The results indicate that multimodal integration using the Late Fusion method effectively leverages the strengths of each modality, resulting in more accurate and robust emotion detection compared to unimodal approaches. Furthermore, the developed system is capable of performing real-time emotion classification through a camera and microphone, making it potentially applicable in various domains, including education, customer service, human-computer interaction systems, and user emotional state monitoring. The primary contribution of this research is the development of a multimodal emotion recognition system that integrates a CNN-BiLSTM-Attention model for the audio modality and MobileNetV2 for the visual modality using a late fusion method. The proposed system leverages emotional information from both modalities simultaneously to enhance the accuracy and stability of real-time emotion detection, thereby offering an alternative solution for the development of AI-based human-computer interaction systems.

Implementasi SIJAPRI Berbasis Website sebagai Solusi Integrasi dan Pengelolaan Data Ketenagakerjaan

2026-07-28T06:56:43+07:00

The Department of Industry and Manpower of Badung Regency still faces challenges in managing job seeker, company, and job vacancy data that are not yet integrated. Manual administrative processes reduce service efficiency and delay the preparation of employment reports. This study aims to develop the Badung Productive and Independent Job Ready Information System (SIJAPRI), a web-based information system designed to integrate employment services while evaluating its functionality and usability. The system was developed using the Waterfall method, which consists of the requirements analysis, system design, implementation, testing, and maintenance phases. System functionality was evaluated using Black Box Testing, while usability was assessed using the USE Questionnaire. The results indicate that SIJAPRI successfully integrates the management of job seeker data, company data, job vacancies, job applications, recruitment selection processes, and employment reporting. All system functions operated according to the specified requirements based on the Black Box Testing results. The usability evaluation using the USE Questionnaire achieved a score of 87.36%, which falls into the very good category. This research contributes by developing a website-based information system known as SIJAPRI (Badung Siap Kerja Produktif dan Mandiri), which integrates job seeker services, companies, recruitment processes, and employment reporting into a single platform; it also provides a functional evaluation using Black Box Testing and a usability assessment using the USE Questionnaire to serve as a foundation for developing digital-based employment services within government agencies.

Pengembangan Sistem Incident Management Web Menggunakan Waterfall Berdasarkan ITIL 4 untuk Meningkatkan Layanan TI Kampus

2026-07-28T06:57:59+07:00

Information technology (IT) services in higher education require consistent and standardized incident management; however, the Incident Management practices at the Information Technology and System Institute (LTSI) of STT Terpadu Nurul Fikri remain suboptimal. This is caused by inconsistent documentation and the absence of standards for incident logging, classification, and prioritization, resulting in an unstructured resolution process. This research aims to implement IT Service Management (ITSM) based on the ITIL 4 framework, particularly the Incident Management practice, to improve the quality of IT services at LTSI STT Terpadu Nurul Fikri. This study employed the Action Research method integrated with the Waterfall method for its software development. Data collection techniques included interviews, observation, and documentation studies, which were then analyzed using Gap Analysis and the Miles, Huberman, and Saldaña qualitative data analysis model. The results of the needs analysis were implemented into an Incident Management system built on the Laravel 12 framework. The system was subsequently evaluated through Black Box Testing and User Acceptance Testing (UAT) involving LTSI Admins and the academic community. The findings indicate that the implementation of ITIL 4 Incident Management produced a more structured service process through automated ticket logging, prioritization, resolution-status monitoring, and the establishment of a Knowledge Base/Known Error Database (KEDB). This is evidenced by the system functionality testing (Black Box Testing) achieving a 100% success rate, and the user acceptance testing (UAT) obtaining a 100% feasibility score, proving that the developed system is capable of enhancing IT service governance on campus.

Perancangan Sistem Early Warning Kebakaran F-GUARD Menggunakan Decision-Level Sensor Fusion Berbasis Mikrokontroler ESP8266

2026-07-29T23:07:42+07:00

Domestic fire is a fatal disaster threat that requires a responsive and reliable early detection system to minimize material losses. This research focuses on the design of the F-GUARD fire early warning system which uses a decision-level sensor fusion approach based on the NodeMCU ESP8266 microcontroller. This system integrates the MLX90614 contactless temperature sensor for precision thermal radiation detection, the MQ-2 smoke sensor, and the Infrared Flame Sensor to detect the presence of fire directly. The novelty of this research lies in the implementation of a single exponential smoothing algorithm to dampen noise in temperature data and a hold time mechanism for 15 seconds on the smoke sensor to eliminate chattering and false alarms. Environmental data is processed in real-time and visualized on an OLED screen, with buzzer actuator management based on non-blocking execution. The system transmits telemetry data to a cloud database every 5 seconds and sends instant emergency notifications via Telegram Bot. The main contribution of this research is the integration of the single exponential smoothing algorithm and hold time mechanism which successfully increased fire detection accuracy to 95% and reduced false alarms by up to 80% compared to conventional systems. Test results show that this multi-sensor integration with digital filtering logic is able to detect fire with an accuracy level of 95% and the average system response speed from detection to notification delivery is recorded at 2.5 seconds. Thus, F-GUARD is proven effective in providing more reliable, stable, and adaptive fire mitigation management for modern household environment protection needs.

Sistem Pengambilan Keputusan Penerimaan Beasiswa KIP Menggunakan Algoritma Fuzzy Mamdani

2026-07-28T06:48:20+07:00

The selection process for KIP Scholarship recipients at UINFAS Bengkulu is still conducted using conventional procedures, resulting in a relatively time-consuming process and potential subjectivity in decision-making. This study aims to develop a web-based Decision Support System using the Fuzzy Mamdani method to assist in selecting scholarship recipients in a more objective and systematic manner. The system was developed using the Waterfall model, which consists of requirements analysis, system design, implementation, testing, and maintenance. Eligibility assessment was based on four criteria: parents' income, number of family dependents, cumulative grade point average (GPA), and academic achievement, which were processed through fuzzification, IF–THEN rule-based inference, and centroid defuzzification. The main contribution of this study lies in the integration of the Fuzzy Mamdani computational model, which specifically combines financial criteria and student academic information to reduce ambiguity in scholarship selection parameters. The implementation results showed that the system produced a defuzzification value of 80 for the test data, indicating that the applicant was categorized as Eligible for the KIP Scholarship. Functional testing using the black-box testing method on seven test scenarios achieved a 100% success rate, indicating that all system functions operated as expected. The results demonstrate that the proposed system is capable of supporting the KIP Scholarship selection process in a more objective, transparent, efficient, and systematic manner compared with the previous manual process.

Rancang Bangun Asisten Virtual untuk Layanan Bengkel Berbasis Retrieval-Augmented Generation dan Tool Calling

2026-07-30T14:46:01+07:00

The development of Artificial Intelligence (AI) has created new opportunities for Small and Medium Enterprises (SMEs) to improve customer service quality through communication automation. C Maestro Workshop in Semarang still handles customer inquiries manually through telephone and messaging applications, resulting in delayed responses to service cost inquiries, service reservations, and vehicle repair status updates. These services also depend on staff availability and cannot be provided consistently on a 24-hour basis, leading to inefficient customer service and potentially reducing customer satisfaction. This study aims to design and develop a Retrieval-Augmented Generation (RAG)-based virtual assistant utilizing the OpenRouter Application Programming Interface (API) to address these issues. The system was developed using a TypeScript-based client-server architecture running on the Node.js runtime environment. The RAG approach was implemented through a context injection mechanism by integrating the workshop's internal knowledge base, which includes service information, estimated service costs, Frequently Asked Questions (FAQ), initial diagnostic guidance, and other supporting information, into the system prompt of the Large Language Model (LLM). In addition, the system implements a tool-calling mechanism that enables the virtual assistant to perform business functions such as service reservations, vehicle repair status inquiries, and complaint escalation to mechanics. Operational data are stored in an SQLite database, while the user interface is provided through a web-based chat application with Server-Sent Events (SSE) support to deliver real-time responses. The evaluation results indicate that the system is capable of generating relevant responses based on the knowledge base, automating the service reservation process, and escalating requests beyond its capabilities. This study contributes by developing a virtual assistant architecture that integrates Retrieval-Augmented Generation (RAG) based on context injection with a tool calling mechanism using the OpenRouter API to support both information services and workshop operational processes. Therefore, the developed virtual assistant offers an effective, practical, and cost-efficient digital solution for supporting the digital transformation of customer services in small and medium-sized automotive workshops.

Pengembangan dan Validasi Sistem Pemantauan Mikroklimat Real-Time Berbasis IoT sebagai Pendukung Evaluasi Kondisi Pengeringan Pascapanen pada Solar Dome Dryer

2026-07-30T14:48:02+07:00

Continuous monitoring of temperature and humidity is necessary to evaluate the microclimate conditions that affect the drying performance of agricultural products in the Solar Dome Dryer. However, manual monitoring has not been able to provide real-time data, while previous research has largely focused on the implementation of devices without validating the measurements against standard measuring instruments under operational drying conditions. This research aims to develop a real-time microclimate monitoring system based on the Internet of Things to support the evaluation of post-harvest drying performance on the Solar Dome Dryer. The system was developed using the DHT22 sensor, ESP32 microcontroller, Wi-Fi connection, and cloud platform to display temperature and humidity data through a web-based and smartphone dashboard. Testing was conducted on the functions of acquisition, transmission, data visualisation, and sensor accuracy by comparing the measurement results with the standard measuring instrument UNI-T UT333. During the testing, the system was able to record and transmit microclimate data stably within the temperature range of 29.1–65.6 °C and relative humidity of 13.7–79.5% RH. In a limited test of 100 transmission cycles with a 10-minute interval, all data were successfully received by the server under network conditions during the testing. Validation using 50 pairs of data resulted in an MAE of 0.150 °C and an RMSE of 0.159 °C for temperature, as well as an MAE of 0.286% RH and an RMSE of 0.293% RH for humidity. The MAPE values were 0.34% and 0.92%, respectively. The contribution of this research is the development of a low-cost microclimate monitoring framework that integrates acquisition, transmission, storage, real-time visualisation, and statistical sensor validation under the operational conditions of the Solar Dome Dryer. The system can be used as a data source to support drying condition evaluations, with the limitation that accuracy validation above 60 °C has not yet been conducted using appropriate reference instruments.

Perbandingan Gated Recurrent Unit dan Time Series Transformer untuk Prediksi Kabut Menggunakan Sliding Window

2026-07-30T14:49:30+07:00

Fog is one of the most hazardous weather phenomena for aviation operations. Dense fog can reduce visibility to below 1,000 meters, potentially causing flight delays, cancellations, and even aviation incidents. To date, fog prediction, particularly at Budiarto Airport, still relies on manual analysis by weather forecasters, making it prone to subjectivity and delays in information delivery. This study proposes and compares two deep learning architectures: the Gated Recurrent Unit (GRU) as an efficient recurrent model, and Time Series Transformer (TST) based on self-attention as a state-of-the-art model for METAR (Meteorological Aerodrome Report) data-based fog event prediction. The METAR data is initially processed using a sliding window technique before becoming a ready-to-use dataset. The dataset comprises 153,838 METAR records from the Budiarto–Curug Meteorological Station spanning from September 2015 to February 2026, which were processed through a METAR code parsing pipeline, BMKG rule-based median imputation, Min-Max normalization, and the construction of a 9-1 sliding window dataset. Experimental results on the test data demonstrate that TST 9-1 delivers the best performance with a Root Mean Squared Error (RMSE) of 0.452427 and a three class classification accuracy (No Fog, Light Fog, Dense Fog) of 88.21%, significantly outperforming GRU 9-1, which achieved an RMSE of 0.883981 and an accuracy of 72.97%. The main novelty of this research lies in the comparative study of GRU and TST architectures for METAR based fog prediction at airports, combined with a sliding window technique and the conversion of visibility regression into a multi class classification of fog events. This research contributes a fog prediction modeling framework capable of processing time-series data sequentially and more effectively, which can serve as a foundation for the development of an accurate, automated early warning system for fog events in airport environments.

Evaluasi Arsitektur Long Short-Term Memory untuk Klasifikasi Gestur Tangan Dinamis Berbasis MediaPipe Sebagai Kendali Presentasi

2026-07-30T14:51:07+07:00

Advances in Human-Computer Interaction (HCI) are driving the use of more natural interaction methods through hand gesture recognition technology. Dynamic gestures have an advantage over static gestures because they contain temporal information such as direction and movement patterns hat is more representative in conveying a command. However, the development of dynamic gesture recognition models still faces challenges in maintaining generalization capabilities for new users, necessitating an evaluation scheme capable of measuring model performance more representatively. This study aims to develop a dynamic hand gesture classification model for controlling PowerPoint presentations based on MediaPipe Hands, MediaPipe Pose, and Long Short-Term Memory (LSTM). The research stages included video data collection, hand and shoulder landmark extraction, data normalization using the midpoint of both shoulders as a reference point, training of eight variations of the LSTM architecture, and evaluation using the Leave-One-Subject-Out Cross-Validation (LOSO-CV) scheme, such that each participant took turns serving as test data to evaluate the model’s generalization ability toward new users. Test results show that the Baseline 1 (B1) architecture delivers the best performance with an average accuracy of 95.77%, precision of 96.03%, recall of 95.90%, and an F1-score of 95.84%, Analysis of the confusion matrix shows that most gestures were correctly recognized, while misclassifications occurred primarily in the “idle” class and for some gestures with similar hand poses. The results of the study indicate that the combination of MediaPipe and LSTM is capable of building a dynamic hand gesture classification model that maintains consistent classification performance in cross-subject testing.

AI-Assisted Labeling for Indonesian Hadith Classification using Four Thematic Categories: Evaluating TF-IDF+SVM and IndoBERT

2026-07-30T14:52:39+07:00

Hadith is the second primary source of Islamic law after the Qur'an, and its large volume makes manual thematic classification time-consuming and inefficient. This study proposes an AI-assisted labeling approach to construct an Indonesian thematic hadith dataset and evaluates the performance of two text classification methods, namely TF-IDF + Support Vector Machine (SVM) and IndoBERT. The dataset consists of 6,600 Indonesian-translated Sahih Bukhari hadiths collected from the Hadith API and categorized into four thematic classes: aqidah, ibadah, akhlak, and muamalah. The annotation process employed Gemini 2.5 Flash with a structured prompt and JSON-based output format, followed by validation performed by a hadith researcher validation on a randomly selected 5% sample, achieving an overall agreement of 73.3%. The annotated data were divided into training and testing sets using an 80:20 stratified split. Model performance was evaluated using Accuracy, Macro F1-score, and Weighted F1-score. Experimental results show that TF-IDF + SVM achieved an Accuracy of 73.1%, a Macro F1-score of 70.1%, and a Weighted F1-score of 73.0%, while IndoBERT achieved an Accuracy of 72.3%, a Macro F1-score of 69.7%, and a Weighted F1-score of 72.2%. The results indicate that the conventional TF-IDF + SVM approach slightly outperformed the Transformer-based IndoBERT model on the proposed dataset. The main contributions of this study are the construction of an Indonesian thematic hadith dataset through AI-assisted labeling and a comparative evaluation of conventional and Transformer-based methods for Indonesian hadith classification.

Rancang Bangun Sistem IoT Monitoring Pemakaian Daya Listrik dan Air Mesin Cuci Laundry Komersial

2026-07-30T14:53:56+07:00

Monitoring electricity and water consumption of washing machines in laundry businesses poses a distinct challenge for laundry owners. This research aims to develop an IoT-based monitoring system for electrical power and water consumption on washing machines using an ESP32, a PZEM-004T sensor, and a YF-S201 water flow sensor. The data captured by the sensors was stored in a MySQL database server via the MQTT protocol and displayed on a monitoring web page built with CodeIgniter, as well as on a Virtuino IoT Android application. The results showed that the electrical sensor was capable of measuring electrical quantities with a reading error rate of 0.5–1% compared to a digital kWh meter. Similarly, the water flow sensor was able to measure water volume with an error rate of approximately 0.2%. The web-based monitoring dashboard can display data and graphs of electricity and water usage, enabling laundry owners to perform analyses for decision-making in future business development. Sensor data storage from the ESP32 to the MySQL server was successfully executed in real-time with a 1-second time delay. Thus, this IoT implementation successfully provides a real-time and accurate operational monitoring system for washing machines.

Implementasi E-Commerce Berbasis Laravel untuk UMKM Lissiger dengan Desain Mobile-Responsive

2026-07-30T14:57:26+07:00

The rapid advancement of digital technology has encouraged Micro, Small, and Medium Enterprises (MSMEs) to adopt electronic commerce in order to improve their business competitiveness. Lissiger MSME, located in Bandar Lampung and engaged in the production of traditional Lampung siger crowns, still relies on conventional marketing and manual transaction recording, resulting in limited market reach and inefficient administrative processes. This study aims to design and develop a mobile-responsive ecommerce platform to expand digital marketing reach and automate transaction management for Lissiger MSME. The system was developed using the Agile Scrum methodology through three structured sprint cycles. Laravel framework, Bootstrap 5, and MySQL were utilized to develop the application System evaluation was conducted using Black Box Testing to verify system functionality and User Acceptance Testing (UAT) involving 10 respondents, consisting of one MSME owner and nine prospective customers, to evaluate user acceptance.The developed platform provides a 50% Down Payment (DP) transaction mechanism and an interactive visualization feature presenting nine stages of siger production as a cultural educational medium. The Black Bo Testing results showed a 100% success rate across 11 functional test scenarios, while the UAT achieved an average score of 88.70%, indicating that the system is highly acceptable for users. These findings demonstrate that the proposed platform effectively supports digital marketing expansion and transaction management for Lissiger MSME. The main contribution of this study is the development of a mobile-responsive e-commerce platform that integrates a 50% Down Payment (DP) payment mechanism aligned with MSME business processes and an interactive visualization of siger production as a medium for cultural education and preservation within a single system.

Analysis of Student GitHub Repository Activity Patterns in Web Framework Programming Courses using K-Means Clustering

2026-07-30T14:44:25+07:00

Instructors in project-based web framework courses frequently lack objective mechanisms to monitor individual student development progress, particularly when engagement occurs asynchronously across multiple weeks and repository activity is not systematically analyzed. This study aims to identify distinct behavioral engagement profiles among students using GitHub repository activity data, and to demonstrate the utility of unsupervised machine learning as a scalable progress monitoring tool for project-based programming courses. The K-Means Clustering algorithm was applied to analyze repository activity patterns of 73 students enrolled in a Web Framework Programming course using Laravel at a vocational higher education institution. Five behavioral features were extracted from each student's GitHub repository, namely total_commit, active_days, avg_commit_per_day, weekend_commit, and last_commit_gap. Following data normalization using StandardScaler, the optimal number of clusters was identified as k=3 using the Elbow Method. The clustering analysis revealed three distinct behavioral profiles: Cluster 0 (51 students, 69.86%) as Passive Learners characterized by low commit activity and a high last_commit_gap indicating deadline-driven development behavior; Cluster 2 (20 students, 27.40%) as Productive Learners demonstrating substantially higher commit intensity and broader repository engagement; and Cluster 1 (2 students, 2.74%) as Highly Consistent Learners exhibiting stable, multi-session repository interaction throughout the project period. As an initial validation of clustering quality, the Silhouette Score of 0.4041 confirms a moderate yet meaningful partition structure within the dataset. The primary contribution of this study lies in demonstrating that mandatory GitHub repository submissions, already required in most project-based programming courses, can be repurposed into an objective, low-cost behavioral monitoring instrument without additional data collection burden. This contributes a replicable, repository-based learning analytics framework that enables instructors to objectively classify student project engagement, supporting early instructional intervention and more process-oriented assessment strategies in software engineering education.

Penerapan Sistem Pakar untuk Diagnosis Kerusakan Smartphone Menggunakan Metode Forward Chaining Berbasis Android

2026-08-01T11:23:30+07:00

This study aims to design and develop an Android-based expert system application used to diagnose smartphone malfunctions using the Forward Chaining method. The main issue in this study is the limited knowledge of lay users in understanding the symptoms and types of smartphone malfunctions before seeking repairs. Research data was obtained through interviews with expert smartphone technicians at CarlCare Mobile Phone Repair Service in Medan, which is an official service center for Transsion vendors, namely Infinix, Itel, and Tecno. The data collected consists of 45 symptoms, 37 types of damage, 44 diagnostic rules, and 37 brief repair solutions. This application was developed using Flutter and Dart, with Firebase as the primary database and SQLite as the local database to ensure the application remains usable offline after synchronization is complete. The Forward Chaining method was applied by matching symptoms selected by the user with IF-THEN rules in the knowledge base to generate a diagnostic conclusion. The implementation results show that the app can display fault categories, symptom lists, diagnostic results, brief solutions, diagnostic history, as well as database synchronization and reset features. Black-box testing results indicate that all features function as expected. Thus, this app can help users obtain an initial diagnosis of Transsion smartphone faults in a simple, fast, and targeted manner.

Penerapan Naive Bayes untuk Klasifikasi Opini Fans Manchester United pada Media Sosial

2026-08-01T11:29:28+07:00

This study examines the application of the Naïve Bayes algorithm to classify the opinions of Manchester United fans in Indonesian-language comments on YouTube. The diverse linguistic forms found in the comments such as slang, abbreviations, jokes, and sarcasm make manual analysis inefficient and potentially subjective. Data was collected from eight YouTube videos using the YouTube Data API v3 and stored in a MySQL database. Of the 4,886 comments obtained, 1,962 were identified as being in Indonesian. A total of 1,000 comments were used as ground truth, consisting of 400 positive, 300 negative, and 300 neutral comments. The data was divided into 80% training data and 20% test data using a stratified split. The processing stages included text preprocessing, TF-IDF weighting, Naïve Bayes classification, and evaluation using a confusion matrix, accuracy, precision, recall, and F1-score. The test results yielded an accuracy of 92.50%, a macro precision of 92.81%, a macro recall of 92.36%, and a macro F1-score of 92.57%. Of the 1,944 comments successfully classified, positive sentiment dominated at 47.58%, followed by negative at 28.34% and neutral at 24.07%. The web-based system, built using Laravel, PHP, and MySQL, is capable of integrating the processes of data extraction, labeling, classification, evaluation, and result visualization.

Evaluasi Efektivitas Strategi Promosi Perguruan Tinggi Menggunakan Algoritma DBSCAN dan Sistem Informasi Geografis

2026-08-01T11:36:55+07:00

Campus promotions that lack a clear direction often lead to inefficient budget use and imbalanced differences in the number of applicants between regions. This study aims to examine how effective the promotional methods used by the University of Muhammadiyah Bengkulu are, using a combination of the DBSCAN algorithm and Geographic Information Systems. The DBSCAN algorithm is used to group data from prospective students based on the density of applicants and identify areas that are unusual or have irrelevant data, while GIS helps to visually map the distribution patterns of these data. The evaluation results show that applicants tend to gather in certain areas that support the success of the promotion, while also identifying potential areas that have not been optimally reached, the model evaluation produces a Silhouette Score of 0.159, which indicates spatial overlap, GIS visualization successfully provides a clear picture of the distribution of students, The contribution of this study is to integrate the DBSCAN algorithm with Geographic Information Systems to evaluate the effectiveness of promotional strategies based on the spatial distribution of prospective students so that it can provide recommendations for more targeted and data-based promotional areas. These insights provide a strategic foundation for designing more efficient, targeted, and data-driven marketing campaigns in high-potential regions.

Perbaikan Postur Kerja menggunakan Metode REBA pada Area Assembly Engine

2026-08-06T10:19:48+07:00

This study aims to analyze ergonomic risks and propose improvements to working posture using the Rapid Entire Body Assessment (REBA) method in the Assembly Engine area. The research was conducted through direct observation, documentation of operators’ working postures, interviews, and distribution of Nordic Body Map (NBM) questionnaires. The results showed that several work activities had moderate to high ergonomic risk levels caused by bending postures, repetitive movements, and non-ergonomic component handling activities. The most common complaints experienced by operators were related to the neck, back, waist, and wrists. Based on the REBA analysis results, several improvements are required, including workstation layout adjustments, use of assistive devices, improvement of working methods, and implementation of ergonomic training. The proposed improvements are expected to reduce the risk of Musculoskeletal Disorders (MSDs), improve operator comfort, and support productivity and occupational safety sustainably within the production environment.

Rancang Bangun Dashboard Web Real-Time Berbasis WebSocket dan MQTT untuk Pemantauan Kualitas Air dengan Fitur Predictive Maintenance

2026-08-06T10:29:51+07:00

Conventional water quality monitoring systems face two main obstacles: high data transmission latency caused by HTTP polling mechanisms and the inability to proactively notify operators of sensor conditions before failure occurs. This study presents the design and implementation of an integrated real-time web dashboard that combines Internet of Things (IoT) architecture with a WebSocket-based interface for water quality monitoring and predictive maintenance notifications. An ESP32 microcontroller acquires pH, turbidity, temperature, and TDS parameters and publishes them asynchronously via MQTT protocol to the HiveMQ public broker. A FastAPI backend subscribes to the broker, processes time-series data streams, and broadcasts telemetry and inference status to a React 18 frontend through persistent WebSocket channels. As a supporting inference component, a 1D-CNN model is adopted as a standard classification engine to distinguish Normal, Degraded, and Critical sensor conditions—complementing a TDS-based Rejection Rate formula used as an aggregate filtration performance indicator. System testing shows an average end-to-end latency of 1.38 seconds from sensor acquisition to dashboard rendering, with 1–2 second data reception intervals confirming WebSocket channel reliability. Conditional rendering on the React interface successfully visualizes three dynamic operational status levels based on Rejection Rate values: Fit for Consumption (green, RR 94.8%), Warning (orange, RR 91.6%), and Not Fit for Consumption (red, RR −79.8%). The resulting platform demonstrates that integrating low-latency message brokering with a high-performance asynchronous server can deliver responsive visualization and proactive maintenance notifications without dedicated GPU hardware. The primary contribution of this study is the development of an integrated monitoring dashboard architecture that combines low-latency messaging protocols with a lightweight inference model, thereby enabling real-time preventive maintenance notifications to be executed instantly on an interactive web interface.