Jurnal Ilmu Komputer dan Sistem Informasi

Analisis Penerapan Internasional Ship and Port Facility Security (ISPS) Code terhadap Peningkatan Efektivitas Keamanan di dermaga Internasional Bulk Liquid (IBL) Pelabuhan Gresik

2026-04-26T18:14:12+07:00

Indonesia's strategic ports face rising maritime threats like piracy and terrorism, necessitating effective ISPS Code implementation. This study analyzes the application of ISPS Code (X) and its impact on security effectiveness (Y) at the International Bulk Liquid (IBL) Terminal, Gresik Port. Employing a quantitative descriptive-explanatory approach, data were collected via Likert-scale questionnaires from 50 purposive respondents (security personnel, PFSO, operators) during January-June 2025, supported by observations and secondary documents. Instruments were validated (r-hitung > 0.235; sig. <0.05) and reliable (Cronbach's α X=0.848, Y=0.804), analyzed using simple linear regression in SPSS. Results show excellent ISPS implementation (mean 4.46) and security effectiveness (mean 4.68), with significant positive influence (t=11.852 > 1.677; sig=0.000; R²=0.745; Y=17.071+0.641X). In conclusion, ISPS Code substantially enhances port security, recommending intensified training and technology integration.

Penerapan Metode K-NN untuk Klasifikasi Pencari Kerja pada Dinas Tenaga Kerja

2026-04-17T16:16:23+07:00

The Pasuruan Manpower Office is one of the agencies that tries to assist companies in the selection process for job seekers who match the criteria for available job vacancies. A common problem in the Job Placement section is the inefficiency in determining suitable job classifications for job seekers. In addition to the lack of human resources in the job selection process, there is also no system that helps the workforce sector in selecting suitable jobs for job seekers. The purpose of this study is to design and analyze a system that can assist the Pasuruan Regency Manpower Office in finding Fields / Classes in Placement according to the criteria of job seekers. The job seeker criteria used are Education Level, Work Experience, Language Proficiency, Last Educational Grade, Major, Expertise, and Training Certificate. The method used in this study is the K-NN method. The K-NN method is a method for classifying objects based on learning data that is closest to the object, by selecting the appropriate K data can be classified based on the nearest neighbor. From the results of the tests carried out 12 times on the testing system against 50 data with training data parameters of 25 data to 50 data and the K values used were 1, 2, 3, 4, 5 and 6, the average classification had an accuracy value of 81.17% and an error value of 18.87%.

Sistem Penetasan Artemia dan Pemberian Pakan Burayak Ikan Guppy Otomatis Berbasis Mikrokontroler

2026-04-21T09:28:53+07:00

Guppy fish (Poecilia reticulata) cultivation requires special attention, especially in terms of feeding the fry, which must be done regularly and consistently. One of the best feeds for fry is Artemia, but the process of hatching and feeding manually often makes it difficult for farmers. This study aims to design and implement an automated system that is able to hatch Artemia and provide food to guppy fry on a schedule using an Arduino Mega 2560 microcontroller and NodeMCU ESP8266. The system is equipped with various sensors such as DS18B20, DHT11, and ultrasonic, as well as actuators such as servo motors, DC motors, and DC water pumps. Feed scheduling is managed using an RTC module and can be monitored through the Telegram interface in real-time. Test results show that the system can hatch Artemia semi-automatically within 24 hours, where the operator still presses a button or starts each cycle, but the movement, position and speed are automatically controlled by the servo and controller, thus providing feed automatically and on time three times a day, and sending status notifications via Telegram with a fast response. This system has been proven to increase efficiency and accuracy in the guppy fish cultivation process and provides a practical automation solution for farmers. This study uses the Research and Development (R&D) method with a systems engineering approach. This method was chosen because it is suitable for designing and developing an Artemia hatching system and automatic feeding. This study designed and implemented a semi-automatic Artemia hatching system that can hatch Artemia eggs and distribute them as food for guppy fry on a scheduled and automatic basis. The system is also equipped with remote monitoring and control features via a Telegram Bot based on the ESP8266 microcontroller and Arduino Mega.

Studi Eksperimental: Efektivitas AI Dalam Meningkatkan Aksesibilitas Informasi PCOS Via Mobile App

2026-05-30T12:16:33+07:00

Polycystic Ovary Syndrome (PCOS) is a hormonal disorder commonly experienced by women of reproductive age; however, the level of literacy regarding PCOS remains relatively low. Limited access to clear and easily understandable information has contributed to a lack of awareness of this condition. This study aims to develop an Artificial Intelligence (AI)-based mobile application called OVAI as an educational medium to improve the accessibility of PCOS information. The research was conducted by distributing questionnaires to female respondents aged 18–45 years using a Likert scale. The collected data were analyzed to determine the level of user acceptance and the consistency of the research instrument. The results showed that the OVAI application was successfully developed with interactive chatbot features and user-friendly educational content. The application was considered helpful in enabling users to better understand information related to PCOS and to access information more practically. Therefore, the OVAI application can be utilized as a digital educational medium for women of reproductive age.

Hardening Keamanan Server eOffice Apache dengan TLS 1.3 dan Fail2ban

2026-05-04T16:25:16+07:00

The security of campus digital services has become increasingly critical due to the rising intensity of automated attacks such as brute-force attempts, vulnerability scanning, and file upload exploitation targeting web-based administrative systems. The eOffice server of Universitas HKBP Nommensen, which serves as the central platform for document management and official correspondence, is also exposed to such threats. This study aims to enhance server security by implementing a defense-in-depth hardening strategy on Apache 2.4. The methodology includes the activation of TLS 1.3 for modern encrypted communication, the implementation of OWASP-compliant security headers, directory isolation to restrict malicious file execution, and the deployment of Fail2ban as a log-based Intrusion Prevention System (IPS) using a multi-jail approach. Evaluation was conducted using SSL Labs, SecurityHeaders.com, and attack log analysis. The results demonstrate significant improvements, highlighted by an upgrade in SSL rating from grade B to A+ and an increase in Security Headers rating to Grade A. In addition, the implemented IPS proved effective in detecting and mitigating automated attacks in real time. In conclusion, the combination of Apache hardening, modern TLS configuration, and log-based intrusion prevention significantly enhances the resilience of eOffice services and can be readily replicated by other institutions with limited resources.

Model Optimasi Alokasi Dana Investasi Terbatas Menggunakan Pendekatan Riset Operasi dan Orange Data Mining

2026-05-07T07:34:27+07:00

Limited capital is a major obstacle in investment decision-making, especially for novice investors. This study aims to develop an optimization model for investment fund allocation with limited capital using an operations research approach (linear programming) and its implementation through Orange Data Mining. A case study was conducted on three digital investment instruments (stocks, mutual funds, gold) with a total capital of IDR 100 million and a maximum risk limit of 10%. The method used is linear programming solved using the Simplex method and Orange Data Mining. The results showed an optimal solution of 25% stocks, 50% mutual funds, 25% gold with an expected return of IDR 10.75 million/year. Method comparison showed identical accuracy (100%) with Orange computation time of 0.3 seconds vs 15 minutes manually. Sensitivity analysis revealed that increasing risk tolerance from 10% to 15% increases returns to IDR 12.5 million. This research contributes to the development of investment decision support tools that are easy to use by novice investors as well as teaching materials for Operations Research.

Analisis Kejelasan Tujuan Pendidikan Remaja Berbasis Machine Learning dan Data Digital Psikometrik

2026-05-02T10:51:48+07:00

This study aims to analyze adolescents’ educational goal clarity based on the integration of psychometric and digital activity data using a machine learning approach. The main problem addressed in this study is the limited utilization of digital and psychological data to understand adolescents’ educational goal characteristics. The methodology includes data preprocessing, clustering analysis using the K-Means algorithm to identify hidden patterns, and classification using Logistic Regression and Decision Tree to build predictive models. The results show that respondents are grouped into three main clusters: low, medium, and high, reflecting different levels of digital engagement and educational goal clarity. Using an 80:20 data split, the results show that on the test set (20%), Logistic Regression achieved the best performance with an accuracy of 0.95 and ROC-AUC of 0.996, while Decision Tree provided more interpretable patterns with an accuracy of 0.80. The most influential variables include frequency of information searching, duration of accessing educational content, and clustering results. The study concludes that productive digital behavior significantly contributes to adolescents’ educational goal clarity, and the hybrid machine learning approach is effective in combining pattern analysis and prediction.

Analisis Pengelompokan Tingkat Pemahaman Materi Siswa Berdasarkan Nilai Ujian Menggunakan Algoritma K-Means

2026-05-03T17:49:10+07:00

Student understanding of learning content is an important indicator of educational success. This study aims to group student understanding based on their exam results in mathematics, English, science, social studies, and Arabic using the K-means Clustering algorithm. The data used consisted of 60 rows of student performance data, which were processed and standardized using Google Colab. The number of clusters was limited to two groups (K=2) to categorize students as having a high level of understanding or a basic level of understanding. The results showed that the K-means algorithm successfully identified groups of students with different levels of understanding based on their average exam scores. The group with a high level of understanding achieved an average score of more than 87.4. Teachers can use these Clustering results as a basis for developing more individualized and effective learning strategies for each group of students.

Pengembangan Aplikasi Presensi Berbasis Deep Learning

2026-05-05T09:18:38+07:00

A facial recognition-based attendance system is a modern solution to overcome the weaknesses of manual attendance methods that are prone to manipulation and recording errors. This study uses a deep learning-based attendance application by implementing a Convolutional Neural Network (CNN) using MobileNetV2, VGG16, and ResNet50 architectures optimized for devices with limited resources. The facial dataset was collected independently and went through preprocessing stages, including normalization, resizing, augmentation, and face detection with OpenCV. The model was trained using TensorFlow and Keras on Google Colab with a GPU. It was then evaluated using a confusion matrix, which yielded accurate predictions with a low error rate. A classification report was also conducted, with an accuracy of 0.98, a precision of 1.00, a recall of 1.00, and an F1-score of 1.00, achieving a very high level of performance, indicating no prediction errors. A Flask web-based application was designed to connect the facial recognition model with the user interface, and was tested in real-time to measure the speed and accuracy of attendance. The results show that the CNN-based attendance application is able to provide a safer, faster, and more efficient attendance alternative compared to conventional methods.

Model UTAUT2 dalam Evaluasi Faktor-Faktor yang Mempengaruhi Niat Pembelian pada Konten Affiliate TikTok Shop

2026-05-03T17:32:46+07:00

The growth of TikTok Shop’s affiliate program provides opportunities for affiliates to earn commissions from uploaded content. However, the low purchase conversion rate indicates challenges in shaping BI. This study aims to analyze the factors influencing purchase intention in TikTok Shop affiliate content by adapting a modified UTAUT2 model. This research employed a quantitative approach through a survey of 410 respondents selected using purposive sampling, with the criteria of active TikTok users who had interacted with TikTok Shop affiliate content. The data were analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM) to examine the relationships among variables in the research model. The results show that only PV has a significant effect on BI. Although EE and HT significantly affect HM, they do not directly influence BI. In addition, HM, TR, FC, and PE do not show a significant effect on BI. This study contributes by extending the application of the UTAUT2 model to the social commerce context, particularly TikTok Shop affiliate content, and emphasizes the importance of enhancing PV as the main factor in encouraging BI.

Penerapan Arsitektur U-Net pada Segmentasi Cacat Biji Kopi untuk Optimalisasi Inspeksi Kualitas

2026-05-05T12:18:51+07:00

Manual visual inspection of coffee bean defects remains prone to subjectivity and inconsistency, necessitating a more accurate and efficient approach. This study proposes a deep learning-based coffee bean image segmentation method using the U-Net architecture to detect the presence of defects in coffee beans using a binary segmentation approach. The dataset consists of 300 coffee bean images evenly divided into 150 images of black coffee and 150 images of insect damage. Annotation was performed using a semi-automatic pseudo-labeling method based on Gaussian filtering, absolute difference, and thresholding to generate ground truth in binary mask format. Training data was enriched through augmentation techniques including horizontal flip, vertical flip, rotation, and brightness-contrast adjustment. The model was trained using a combined loss function of Dice Loss and Binary Cross-Entropy with the Adam optimizer over 15 epochs with an early stopping mechanism. Evaluation results demonstrate excellent performance with a Mean IoU of 0.9240, Precision of 0.9707, Recall of 0.9495, and F1 Score of 0.9600, with an overall correct prediction rate of 97.45% based on pixel-level confusion matrix analysis. These results indicate that the U-Net architecture is capable of segmenting defective coffee bean areas accurately and consistently, making it a promising foundation for the development of an automated coffee quality inspection system.

Rancang Bangun Sistem Informasi Enterprise Resource Planning (ERP) Pada Dinas Kesehatan Kota Medan

2026-05-11T08:49:13+07:00

The Medan City Health Office faces various challenges in managing human resources and inventory due to manual administrative processes, resulting in data inaccuracies, delays, and low operational efficiency. This study aims to design and implement an Enterprise Resource Planning (ERP) system using Human Resource Management (HRM) and Inventory Management (IM) modules. The research employed a qualitative method with data collection techniques consisting of observation, interviews, and literature study. System development used the Rapid Application Development (RAD) method, which includes requirement planning, workshop design, and implementation stages. The results indicate that the ERP system is capable of integrating employee management, attendance, payroll, training, inventory, and stock management processes. The developed system improves data accuracy, accelerates administrative processes, and supports more effective decision-making. Therefore, the implementation of ERP at the Medan City Health Office can enhance efficiency, transparency, and the quality of healthcare services.

Analisis Perilaku Adopsi Livin Merchant Pada Pelaku UMKM di Surabaya Raya Menggunakan Model UTAUT2

2026-05-30T11:29:15+07:00

The development of digital technology has encouraged the transformation of financial services in the micro, small, and medium enterprises (MSMEs) sector, one of which is through the use of the Livin’ Merchant by Mandiri application. However, the adoption process of digital services still faces several challenges, such as low technological literacy, risks associated with application usage, and low trust in system security and banking institutions. This study aims to analyze the factors influencing the adoption behavior of Livin’ Merchant among MSME actors in Greater Surabaya using the modified Unified Theory of Acceptance and Use of Technology 2 (UTAUT2) model with the addition of perceived security, perceived risk, institution trust, and technology trust variables. This study employed a quantitative approach through a survey of 521 MSME actors using Livin’ Merchant in Greater Surabaya. The data were analyzed using Partial Least Square-based Structural Equation Modeling (SEM-PLS) with the assistance of SmartPLS 4.0 software. The results indicate that performance expectancy, social influence, hedonic motivation, price value, perceived security, and institution trust have a positive and significant effect on behavioral intention. In addition, facilitating conditions have a positive and significant effect on use behavior, while perceived risk has a negative and significant effect on use behavior. This study concludes that application benefits, social support, system security, and trust in banking institutions are important factors in increasing the adoption of Livin’ Merchant among MSME actors in Greater Surabaya.

Implementasi Metode Goal-Directed Design Dalam Perancangan Antarmuka Website MDMC Kabupaten Klaten Berbasis Kebutuhan Pengguna

2026-04-30T15:50:45+07:00

This research is driven by the need for structured information media, specifically a website, to support disaster response reporting at MDMC Klaten Regency. Previously, information remained fragmented on social media, necessitating a more comprehensive platform. This study aims to design the MDMC Klaten Regency website interface based on user needs to ensure effective information delivery. The method employed is Goal-Directed Design (GDD), consisting of Research, Modeling, Requirement, Framework, Refinement, and Support stages. Data were collected through observations, interviews, and document studies, then transformed into Volunteer and General Public personas. The result is a website interface prototype featuring report publication workflows and location mapping visualization as primary features. Usability evaluation using the System Usability Scale (SUS) instrument with 59 respondents yielded an average score of 87.97. These results place the system in the "Acceptable" category with a "Grade A" rating and an "Excellent" Adjective Rating. In conclusion, the GDD approach is effective in producing interface designs aligned with user goals, characterized by high levels of navigation ease and information clarity. This study recommends further development of map interactivity to continuously optimize the user experience.

Perancangan Antena Mikrostrip Array 2 Elemen untuk Aplikasi 5G Menggunakan Software CST

2026-05-20T14:10:48+07:00

The development of 5G communication technology requires antenna devices that have good performance, small size, and are able to operate at high frequencies. One alternative is a microstrip array antenna because it has advantages in size, ease of manufacture, and the ability to increase gain. This study discusses the design process and performance testing of a microstrip array antenna that functions at a frequency of 5GHz for use in 5G communication systems. The design and testing process were carried out using CST Studio Suite software. The simulation results show that the designed antenna is able to operate at a frequency of 5GHz with a return loss value of less than -10 dB, where the results obtained based on the simulation are -24.42dB. VSWR below 2, where the results obtained based on the simulation are 1.12, and desired bandwidth is wide, which is produced in this design is 380MHz. Thus, the designed two element microstrip array antenna has been successfully designed with the standard parameters specified for use in a 5G network communication system.

Implementasi Smart Home Berbasis IoT: Studi Simulasi Menggunakan Cisco Packet Tracer

2026-05-26T09:27:04+07:00

The rapid development of Internet of Things (IoT) technology enables automatic and remote control of home appliances. However, direct hardware implementation often entails high costs and complex configurations. This study aims to design and analyze the performance of a Smart Home system through a simulation approach using Cisco Packet Tracer software. The method encompasses network topology design, core network service configuration, and functionality and Quality of Service (QoS) testing to validate the proposed design. Simulation results demonstrate that the constructed system operates successfully, with the Smartphone effectively controlling electronic devices in real-time. QoS testing recorded 0% packet loss and an average latency of 27 ms, indicating reliable connectivity between user devices and the IoT Server. This study concludes that simulation using Cisco Packet Tracer is effective as a preliminary validation stage in designing Smart Home infrastructure before physical implementation.

Implementasi Sistem Pemesanan Restoran Berbasis Web Menggunakan Kode QR dan Payment Gateway

2026-05-30T11:31:13+07:00

Kurangnya koordinasi dan efisiensi waktu yang rendah merupakan hambatan utama dalam model layanan restoran tradisional. Untuk mengatasi masalah ini, penelitian ini mengimplementasikan sistem pemesanan dan pembayaran berbasis web dengan memanfaatkan teknologi Kode QR sebagai instrumen digitalisasi transaksi terintegrasi. Pengembangan sistem mengikuti metodologi waterfall, dengan pengujian dilakukan melalui metode pengujian black box. Hasilnya menunjukkan bahwa sistem yang dikembangkan secara efektif memfasilitasi pelanggan dalam melakukan pemesanan sendiri dan mempercepat aliran informasi pesanan kepada kasir dan staf dapur secara real-time. Berdasarkan pengujian fungsional, semua fitur pembayaran baik tunai maupun digital beroperasi sesuai dengan desain. Studi ini menyimpulkan bahwa implementasi sistem berbasis web dan Kode QR secara signifikan meningkatkan efisiensi operasional restoran sekaligus memberikan pengalaman transaksi yang lebih nyaman dan praktis bagi pengguna.

Rancang Bangun Website Travel Umroh Berbasis Web Sebagai Media Informasi dan Digital Trust

2026-05-26T16:24:34+07:00

Umrah travel business is a service sector that strongly depends on the trust of prospective pilgrims. Therefore, service providers need an official information medium that is transparent, easily accessible, and able to validate the company’s identity. This study aims to design and develop a web-based umrah travel website as an information medium and a supporting platform for digital trust at Harira Tour and Travel. The system development method used in this study is the prototype method, consisting of requirement gathering, design, implementation, evaluation, and testing stages. The website was developed using WordPress CMS with several main features, including homepage, company profile, umrah packages, gallery, testimonials, contact information, and WhatsApp integration. The results show that the website is able to present service information in a structured manner, provide responsive access on desktop and mobile devices, and help prospective pilgrims obtain official information more easily. Black box testing on 9 main features shows that all features function according to the functional requirements with a success rate of 100%. This website supports digital trust through company identity, transparent package information, activity documentation, testimonials, and official communication channels.

Prototype Design Of Adaptive Composite Springs As An Anti-Seismic Infrastructure System For Multi-Story Buildings

2026-05-20T13:13:07+07:00

This study aims to design and formulate earthquake dampers composed of carbon fiber and epoxy resin, employing a specialized configuration to produce anti-seismic devices that are corrosion-resistant, adaptively flexible, and capable of supporting building loads effectively. A quantitative design-based methodology was applied, encompassing prototype design, material selection, and performance testing against vertical loads and seismic vibrations. Evaluation was conducted using a hydraulic press to determine maximum vertical load capacity and a shake table to simulate horizontal and vertical seismic activity according to the Richter scale. The results indicate that the composite dampers can absorb seismic energy bidirectionally, maintain structural integrity without significant material degradation, and require minimal maintenance. These findings demonstrate the potential of carbon fiber and epoxy resin-based dampers as adaptive seismic isolation systems that are robust, durable, and suitable for multi-story buildings, while meeting the demands for economical, efficient, and sustainable infrastructure solutions.

Integrasi Algoritma DBSCAN Dengan Sistem Informasi Geografis Untuk Mengidentifikasi Cluster Wilayah Rawan Kebakaran Provinsi Riau

2026-05-08T08:18:17+07:00

Riau Province is one of the regions in Indonesia that is prone to forest and land fires (karhutla), especially during the dry season. Accurately identifying the distribution patterns of fire-prone areas is crucial in supporting disaster mitigation and management efforts. This study aims to identify the spatial patterns (Cluster) of forest and land fire-prone areas (karhutla) in Riau Province using the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm integrated with a Geographic Information System (GIS). This study uses NASA-MODIS data taken from 2020 to 2024 with 840 data records. The analysis results show that DBSCAN is able to effectively group hotspots, with Cluster 2 being the largest cluster covering 297 karhutla points in Bengkalis, Rokan Hilir, and Dumai. The large number of points in this cluster is due to the high frequency of forest and land fires between 2020 and 2024. However, Cluster 7 shows the best density quality with a Silhouette Coefficient value of 0.872, surpassing Cluster 2 which has a value of 0.638. The overall average Silhouette Coefficient value is 0.683, indicating that the cluster modeling is quite optimal. A total of 57 hotspots are categorized as noise, but still provide a picture of the distribution of forest and land fires. GIS-based map visualization reveals that most fire hotspots are located in peatlands and dry vegetation areas that are consistent from year to year. The results of the study confirm that the use of appropriate DBSCAN parameters (epsilon and minPts) produces accurate spatial visualization and supports more effective and targeted mitigation strategies and fire monitoring based on priority areas.

Geo-Insight MBG: A Geospatial Platform to Measure the Impact of the Free Nutritional Meal Program on Children's Nutritional Status

2026-05-30T22:04:12+07:00

The Free Nutritional Meal Program (MBG) is a strategic national policy designed to improve the nutritional status of school-aged children. However, impact evaluation remains largely manual and administrative, hindering the identification of priority areas and integrated monitoring of anthropometric changes. This study aims to develop Geo-Insight MBG, a web-based geospatial platform to measure the program’s impact on children’s nutritional status. The methodology employs the Waterfall software development model and quantitative descriptive analysis of height, weight, and body mass index data collected before and after the intervention. System testing demonstrates that the platform successfully visualizes beneficiary distribution, automatically calculates nutritional categories based on z-scores, and generates interactive dashboards and regional heatmaps. Preliminary results from a pilot sample of 15 children indicate that 73.3% showed nutritional improvement, with an average z-score increase of +0.83. While these findings suggest the platform’s potential for enhancing location-based monitoring and data-driven decision-making, the limited sample size necessitates cautious interpretation of the program’s effectiveness. Future studies with larger, multi-regional cohorts are needed to validate these results. Nonetheless, Geo-Insight MBG demonstrates promise as a scalable digital evaluation tool that could support optimized resource allocation and targeted interventions in national school nutrition programs.

Perancangan Sistem IoT Untuk Monitoring Getaran dan Stress Pada Kanopi Rumah Berbasis ESP32

2026-06-01T17:13:07+07:00

The development of Internet of Things (IoT) enables real-time and continuous monitoring systems. Lightweight structures such as house roofs or canopies are vulnerable to environmental and load changes, yet monitoring is still commonly performed manually, making early detection difficult. This study aims to design and implement an IoT-based canopy structure monitoring system using ESP32 integrated with an MPU6050 vibration sensor, a load cell with HX711 module, and a DHT22 temperature and humidity sensor. Measurement data are transmitted via WiFi to a server for database storage and visualization through a web-based dashboard. Furthermore, the collected data are processed using the Relative Corrosion Potential Index (IPKR), calculated based on normalized parameters including temperature, humidity, vibration, and load variation. The results show that the system is capable of performing real-time data acquisition, transmission, and visualization effectively. The system also provides structural condition indicators based on IPKR values classified into safe, warning, and danger categories. Therefore, the developed system can serve as an effective and informative early monitoring solution for detecting changes in lightweight structure conditions.

Perancangan Sistem Asisten Produktivitas Pribadi Cerdas “Prioritas Bot” Menggunakan Large Language Model (LLM) Berbasis N8N dan Telegram

2026-05-11T14:21:32+07:00

The complexity of task and workflow management in the digital era necessitates intelligent assistive tools. This study designs an Intelligent Personal Productivity Assistant System by integrating three primary components: (1) A Large Language Model (LLM) GPT-4o serving as the processing core for natural language understanding and contextual response generation; (2) The n8n low-code automation platform, acting as an orchestrator to manage workflows and connect the LLM to external APIs such as Google Calendar; and (3) A Telegram Bot utilized as the primary user interface due to its high accessibility. The design methodology employs a Prototyping approach adapted from Pressman & Maxim (2020), enabling iterative development and functional validation. Black Box Testing was applied for functional evaluation, while the System Usability Scale (SUS) was used to measure usability, involving 20 active university students in the Bogor region selected through Purposive Sampling. The research resulted in a functional prototype capable of managing schedules end-to-end via conversational natural language. The system achieved an average SUS score of 88.5 from 20 respondents, classified as Excellent with a Grade A rating, indicating the significant potential of integrating LLMs and workflow automation to effectively enhance personal productivity.

Analisis Aplikasi E-Money Berbasis Mobile untuk Transaksi Pembayaran di Dharmawangsa Foodcourt Menggunakan Metode Usability Testing

2026-05-06T15:05:28+07:00

Dharmawangsa Foodcourt Medan relies entirely on cash-based transactions, generating persistent operational issues including prolonged queuing, cashier miscalculation errors, and the absence of automated financial recording. This condition is inconsistent with the digital literacy profile of its primary users, namely the academic community. This study aims to design, develop, and evaluate an Android-based electronic payment application as a transaction digitalization solution within a campus foodcourt environment. System development employed the Waterfall model, encompassing requirements analysis, system design, implementation, testing, and maintenance phases. Data were gathered through field observation, structured interviews with foodcourt management and tenants, literature review, and transaction workflow documentation. System evaluation was conducted using two approaches: black-box testing to verify complete feature functionality, and usability testing based on the ISO 9241-11 standard involving 10 active users. Black-box testing confirmed that all features operated according to specifications without functional failures. Usability testing produced an overall score of 84.95%, classified as Very Good, comprising effectiveness at 86.00%, efficiency at 83.33%, and user satisfaction at 85.33%. These findings demonstrate that the developed system effectively replaces cash transactions, enhances financial recording transparency, and delivers a faster and more structured transaction experience for users within higher education foodcourt settings.

Implementasi Metode Double Diamond Dalam Perancangan UI/UX Aplikasi Halopet Marketplace Layanan Hewan Peliharaan

2026-05-20T14:05:02+07:00

The rapid growth of the pet care industry in Indonesia has not been accompanied by integrated service systems, resulting in various issues such as limited information, inefficient booking processes, and poor schedule management. This study aims to design the UI/UX of the Halopet application as a pet care service platform using the Double Diamond method to produce user-oriented design solutions. The evaluation was conducted using Heuristic Evaluation and Usability Testing based on the ISO 9241-11 standard. The results show an effectiveness score of 97% and an efficiency score of 97.19%, with a System Usability Scale (SUS) score of 92, categorized as excellent. These findings indicate that the proposed application design is capable of improving usability, efficiency, and user satisfaction in accessing pet care services. This study also contributes by proposing an integrated pet care service platform design through a UI/UX approach focused on user experience and a high level of usability.

Klasifikasi Multi-Label Dan Ekstraksi Entitas Pada Ulasan Aplikasi Blu by BCA Digital Menggunakan IndoBERT

2026-05-20T14:08:00+07:00

The growth of digital banking services in Indonesia has heightened the need to understand factors influencing users' application continuance intention. However, prior studies remain limited to single-label classification and general sentiment analysis, lacking the ability to capture the complexity of information in Indonesian-language user reviews in a structured manner. This study aims to perform multi-label classification based on four Expectation-Confirmation Model (ECM) factors—Confirmation, Perceived Usefulness, E-satisfaction, and Perceived Security—and to extract six Named Entity Recognition (NER) entities from Blu by BCA Digital application reviews using IndoBERT. The dataset was collected from Google Play Store and Apple App Store covering January to December 2025, yielding 3,389 Indonesian-language reviews after filtering. The study employs a single-task approach, applying oversampling and Focal Loss for multi-label classification, and token augmentation with Conditional Random Field (CRF) for NER. Annotation validation using Krippendorff's Alpha yielded average values of 0.856 for intent labels and 0.919 for NER entities. Results show that the best classification model achieved an F1-Score of 0.798 with a Hamming Loss of 0.131, while the best NER model achieved an F1-Score of 0.812. This study demonstrates that IndoBERT is effective for analyzing digital banking application reviews in identifying ECM factors and extracting domain-specific entities, thereby offering potential support for developers in automatically understanding user needs.

Implementasi Algoritma Random Forest untuk Penentuan Relawan Pendonor Darah Potensial pada Palang Merah Indonesia (PMI) Kabupaten Kudus

2026-04-29T09:17:39+07:00

Palang Merah Indonesia (PMI) Kudus Regency faces an imbalance of blood stock across blood types, frequently causing difficulties in meeting blood demand during emergencies. Manual donor data management has been the primary factor behind slow donor recruitment processes. This study develops a machine learning-based prediction system using the Random Forest algorithm to identify blood donor volunteers with a high likelihood of donating again. The system is built as a web application using the CodeIgniter framework and Python, integrated with a Geographic Information System (GIS) via Leaflet.js for spatial visualization of donor distribution. System development follows the CRISP-DM approach, using historical donor data from PMI Kudus Regency as model training material. Evaluation results demonstrate strong model performance with an accuracy of 87.45% and a ROC-AUC of 0.912, while a cross-validation ROC-AUC of 0.897 confirms model consistency without overfitting. The system generates a ranked list of potential donors based on the highest probability values, organized by sub-district and blood type, thereby accelerating PMI officers' response in fulfilling blood demand.

Analisis Sentimen Berita Keuangan Berbahasa Indonesia Menggunakan IndoBERT dan LSTM untuk Klasifikasi Tren Harga Saham pada Lima Emiten Blue Chip LQ45

2026-05-16T07:53:55+07:00

Blue chip stock price movements within the LQ45 index are non-linear and stochastic, causing conventional statistical approaches to fail in generating reliable predictions. This study develops and evaluates a hybrid deep learning architecture combining IndoBERT—a BERT-based language model pre-trained on Indonesian-language corpora—with Long Short-Term Memory (LSTM) networks to classify daily stock price trend direction (up/down). Trend labels are derived from daily return values: a return greater than zero is labeled Up (1), otherwise Down (0). Sentiment scores were extracted via IndoBERT fine-tuning from 9,819 Indonesian-language financial news articles (CNBC Indonesia) and merged with historical OHLCV data and technical indicators as model features. Experiments were conducted on five LQ45 blue chip equities: BBRI.JK, BBNI.JK, BBCA.JK, BMRI.JK (Financial Sector), and GOTO.JK (Technology). The hybrid model outperformed the baseline on three of five equities, with the highest accuracy improvement on BBCA.JK (+23.21 points, from 33.93% to 57.14%) and BMRI.JK (+12.50 points, from 46.43% to 58.93%). The overall average Relative Error Reduction (RER) reached +9.81%, demonstrating that integrating IndoBERT sentiment significantly enhances LSTM-based stock trend prediction in the Indonesian capital market.

Implement’s Steganografi LSB dan Kriptografi AES-256 untuk Pengamanan Data Citra Digital

2026-05-17T18:18:35+07:00

Digital data security has become a critical issue in the digital transformation era, particularly in the exchange of digital images that are vulnerable to interception and manipulation by unauthorized parties. This study aims to implement a layered data security method by combining Advanced Encryption Standard (AES) 256-bit cryptography and Least Significant Bit (LSB) steganography on PNG-format digital images. The approach used is encrypt-then-embed, where the secret message is first encrypted using AES-256 in CBC mode, then the ciphertext is embedded into the cover image pixels through LSB bit substitution. Testing was conducted on five test images with varying resolutions (256×256 to 1024×768 pixels) using Peak Signal-to-Noise Ratio (PSNR), Mean Squared Error (MSE), entropy, avalanche effect, and computation time metrics. The test results show an average PSNR of 70,58 dB and MSE of 0.11, indicating excellent visual quality of the stego images. Security analysis shows an average avalanche effect of 50.29%, approaching the ideal value of 50%, and stego image entropy approaching 7.0545 bits/pixel. The computation time for encryption and embedding processes ranges from 0,005847 to 0,006078 seconds. The combination of AES-256 and LSB has proven effective in providing layered security: AES-256 ensures data confidentiality, while LSB conceals the message existence within the image without significant visual degradation.

Analisis Sentimen Publik Terhadap Revisi UU TNI 2025 Menggunakan Algoritma Naïve Bayes

2026-05-18T16:29:47+07:00

The development of public opinion regarding the revision of the 2025 Indonesian National Armed Forces Law (UU TNI) on social media has generated various responses that are difficult to analyze manually due to the large and unstructured amount of data. This condition requires a computational approach that is able to systematically identify public sentiment trends. This study aims to analyze public sentiment towards the revision of the 2025 TNI Law using the TF-IDF-based Naïve Bayes algorithm and evaluate the performance of the classification model used. The research data was obtained through crawling techniques from YouTube user comments related to the revision of the 2025 TNI Law. The data processing stages include cleaning, case folding, tokenizing, normalization, stopword removal, and stemming before TF-IDF weighting and the classification process using Naïve Bayes. The results of the study of 1826 data show that public sentiment is dominated by the neutral category at 79.8%, while positive sentiment is 13.1% and negative sentiment is 7.0%. The model evaluation yielded an accuracy of 77.11%, but the model showed a bias toward the majority class, resulting in suboptimal classification of positive and negative sentiments. Based on these results, the Naïve Bayes method is quite effective as an initial approach in sentiment analysis, but it still has limitations in handling imbalanced datasets and the complex characteristics of social media language. Therefore, the development of more adaptive methods is needed to improve the quality of sentiment classification results.

Evaluasi Usability Antarmuka Instagram pada Mahasiswa

2026-05-20T18:10:46+07:00

This study examines the usability of Instagram’s interface among university students since Instagram has become one of the social media platforms most frequently used in daily life and academic activities. The study aims to find out how easy and comfortable Instagram is to use based on Nielsen’s usability dimensions which include learnability, memorability, efficiency, error prevention, and satisfaction. This research used a descriptive quantitative method by distributing a 5-point Likert scale questionnaire to 76 active university students in May 2026. The results showed that the overall usability score reached 3.88 and was categorized as “Good”. The satisfaction dimension obtained the highest score of 3.98 while error prevention received the lowest score of 3.75. The findings also revealed that some students were still confused by the similarity between the Reels and Explore features while the placement of advertisements was considered disruptive to navigation. In addition, many respondents suggested adding a study focus mode to support productivity while using Instagram. Overall, Instagram already provides a good usability experience, although improvements are still needed in navigation consistency, error prevention, and features that better support student needs.

Evaluasi Pengaruh Parameter Alpha terhadap Akurasi Metode Single Exponential Smoothing pada Data Persediaan Barang Retail

2026-05-19T07:48:56+07:00

This study aims to evaluate the effect of the alpha parameter on the accuracy level of the Single Exponential Smoothing (SES) method on retail inventory data. The evaluation was conducted using the MAD, MSE, and MAPE error values. The study used Toko Murni's retail inventory data from January 2025 to February 2026, consisting of white rice, cooking oil, bread flour, and 60 ml Bango sweet soy sauce. The evaluation process was carried out using a variation of alpha values from 0.1 to 0.9. The evaluation results showed that low alpha values provide a better level of prediction accuracy than high alpha values. In the white rice data, the use of alpha 0.1 resulted in a MAD value of 75.41, MSE 7167.35, and MAPE 16.86 with a prediction result of 460.43. For cooking oil, alpha 0.1 resulted in a prediction value of 75.66 with a MAPE of 18.6, while for 60 ml Bango sweet soy sauce, it resulted in a prediction of 80.51 with a MAPE of 12.79. Meanwhile, in the bread flour data, the optimal alpha value was obtained at alpha 0.2 with a predicted result of 67.19 and MAPE of 19.71.

Use of Data Visualization Techniques in Bioinformatics for Time-Based Gene Expression Pattern Analysis

2026-05-07T13:37:40+07:00

This study explores data visualization techniques in bioinformatics for analyzing time-series gene expression patterns. It examines how different visualization approaches support the interpretation of large-scale temporal gene expression data. A dataset comprising 4,381 genes across 24 time intervals was analyzed using heatmaps, Principal Component Analysis (PCA), volcano plots, and dendrograms. Heatmaps were used to observe expression correlations, PCA was applied to reduce dimensionality, volcano plots identified differentially expressed genes between conditions, and dendrograms grouped genes with similar expression profiles. The PCA results showed that the first two principal components accounted for 42.32% of the total variance, indicating that these components captured a substantial but not complete portion of the data structure. Volcano plot analysis detected differentially expressed genes based on log2 fold change > 1 and p-value < 0.05, while dendrogram visualization revealed several major clusters with comparable temporal expression patterns. Overall, the findings suggest that combining multiple visualization methods can improve the exploratory analysis of temporal gene expression data by clarifying patterns, highlighting potentially relevant genes, and supporting further biological interpretation. Rather than serving as standalone evidence for clinical application, these visual approaches provide a useful analytical foundation for subsequent validation, biomarker investigation, and large-scale omics research.

Perencanaan Strategis SI/TI Menggunakan Metode Tozer di PT Sinar Sakti Matra Nusantara Bandung

2026-04-30T15:52:43+07:00

PT Sinar Sakti Matra Nusantara (MATRA) is a manufacturing company specializing in machinery services, including turbine repair, mechanical and tooling services, and technical consultation. The main problem is the absence of an integrated information system across business units, causing inefficiencies in operational workflows. The company also lacks a dedicated IT division to oversee IS/IT development. This research aims to: (1) map the competitive business situation of PT Sinar Sakti Matra Nusantara; (2) analyze IS/IT conditions within internal and external environments; and (3) formulate an IS/IT strategy that effectively optimizes business processes. The study applies the Tozer Method, consisting of five phases: Phase 0 (determining context and scope), Phase 1 (identifying business information and support needs), Phase 2 (evaluating system suitability and identifying solutions), Phase 3 (identifying strategic solutions), and Phase 4 (preparing and executing an implementation plan). A qualitative approach was adopted through literature review, interviews, and direct observation. Analytical tools include SWOT, Value Chain, PEST, Five Force Model, Critical Success Factor (CSF), and McFarlan Strategic Grid. The findings reveal that IS/IT development has not been implemented optimally, evidenced by the absence of an IT division and integrated systems. Eight systems and applications are proposed: a Personnel Administration System, Order Information System, Production Information System, Inventory System, Financial Information System, Consumer Order and After-Sales Monitoring Application, Delivery Application, and the company Website. This research concludes that the Tozer Method successfully produced a comprehensive IS/IT strategic roadmap, enabling PT Sinar Sakti Matra Nusantara to grow more competitively alongside rapid technological advancements.

Pengembangan Sistem Informasi Manajemen Pemesanan Berbasis Website di UMKM Nenda Print

2026-06-07T22:00:23+07:00

Micro, Small, and Medium Enterprises (MSMEs) play a vital role in the Indonesian economy, but they often face challenges in managing order and financial data due to manual recording. This study aims to develop a website-based order and profit data recording information system for Nenda Print MSMEs to improve business management efficiency. The system development method includes needs analysis, system design, implementation, testing, and evaluation. The developed website-based system is equipped with features for customer data management, order recording, financial transaction recording, automatic profit calculation, and real-time presentation of reports and graphs. Test results show that the system is able to simplify the data recording process, reduce the risk of data loss, and accelerate sales and profit calculations.

Perancangan Sistem Booking Mini Soccer dengan Payment Gateway dan Dashboard Analitik Menggunakan Metode RAD

2026-05-15T09:34:25+07:00

Information systems play an important role in improving the effectiveness and efficiency of sports facility rental services. S-Soccer Kota Bima still uses a conventional booking system that causes schedule conflicts, delays in confirmation, and unintegrated transaction management. This study aims to design a website-based mini soccer field booking system equipped with a payment gateway feature. The system development applies the Rapid Application Development (RAD) method through the stages of requirements planning, design workshop, and implementation. The system provides features for real-time schedule booking, customer data management, digital payments, and an administrative analytics dashboard. System testing was conducted using the Black Box Testing method on 12 testing scenarios covering registration, login, booking, payment, and data management features. The testing results indicate that all features operated according to user requirements with a success rate of 100%. The developed system is able to improve the efficiency of the booking process, accelerate payment validation, and minimize schedule conflicts compared to the conventional method previously used at S-Soccer Kota Bima.

Klasifikasi Lonjakan Ekstrem Nilai Tukar USD/IDR Menggunakan Gaussian Naïve Bayes

2026-05-26T13:35:40+07:00

The fluctuation of the Rupiah exchange rate against the US Dollar (USD/IDR) potentially triggers macroeconomic instability. This study aims to classify potential extreme surges in the USD/IDR exchange rate using data mining techniques with the Gaussian Naïve Bayes algorithm. A total of 502 daily historical observation data were extracted into four continuous predictor features: volatility, closing difference, upper bound difference, and lower bound difference. The evaluation was conducted using an 80% training and 20% testing data split. The results show that the model can identify "Normal" and "Extreme" classes with an accuracy of 96.04%, a precision of 62.50%, and a recall of 83.33%. The 5-Fold Cross Validation test yielded an average cumulative accuracy of 95.43%, confirming that the model's performance is stable and does not experience overfitting. In conclusion, the Gaussian Naïve Bayes algorithm is proven to be effective and reliable as an early warning system against the risk of extreme foreign exchange rate surges.

Implementasi Metode Waterfall untuk Sistem Penjadwalan Perkuliahan pada Universitas Muhammadiyah Bima

2026-05-15T09:33:37+07:00

Course scheduling is a crucial activity in academic management because it involves organizing lecturers, courses, classrooms, and time allocations in an integrated manner. At Universitas Muhammadiyah Bima, the scheduling process was still conducted manually using spreadsheets, resulting in frequent schedule conflicts, repeated revisions, and inefficient management processes. This study aims to design and implement a web-based course scheduling system capable of overcoming these problems through an automatic constraint-based validation mechanism. The system was developed using the CodeIgniter 4 framework and the Waterfall development method. The main contribution of this research lies in the implementation of real-time automatic validation capable of detecting and preventing conflicts involving lecturers, classrooms, and class schedules before the data are stored in the database, thereby producing a more consistent and reliable scheduling process compared to the previous manual method. System testing was carried out using the Black-box Testing method on 10 testing scenarios covering input validation, schedule conflict detection, data manipulation, and schedule information access. The testing results showed that all scenarios were successfully executed with a success rate of 100%, indicating that all system functions operated in accordance with user requirements. Therefore, the developed system is able to improve the effectiveness of course scheduling management while minimizing scheduling conflicts within the academic environment.

Modifikasi Keamanan Otentikasi OTP Menggunakan Algoritma HMAC-SHA256 pada Sistem Informasi PT Indonesia Gadai Oke

2026-05-30T23:01:20+07:00

The security of user data in web-based information systems is frequently compromised by the weaknesses of conventional authentication mechanisms that rely solely on static passwords. Attacks such as credential theft and brute force on the internal information system of PT. Indonesia Gadai Oke necessitate an additional layer of security to protect sensitive customer data and financial transactions. This research aims to design and implement a Two-Factor Authentication (2FA) security system using the HMAC-SHA256 (Hash-based Message Authentication Code with Secure Hash Algorithm 256-bit) algorithm, integrated with Trusted Device features and Web Push notifications. The applied method is Time-based One-Time Password (TOTP) with a 30-second time interval. A unique code is generated on the server side through a 32-bit dynamic truncation process of the SHA-256 encryption result, which combines a secret key with a timestamp. Web Push Notification was chosen as the distribution medium to eliminate SMS operational costs and minimize delivery latency. System testing was conducted using Black Box Testing and Security Testing methods with a scenario of 10 experimental repetitions. The results indicate that the system achieved a 100% functional success rate in validating authorized users. In terms of security, the system proved effective in mitigating threats with a 100% success rate in rejecting SQL Injection, Cross-Site Scripting (XSS), and Replay Attacks through a single-use token validation mechanism. This implementation successfully reduced the risk of account hijacking and improved the efficiency of the authentication process at PT. Indonesia Gadai Oke.

Analisis Potensi Customer Churn Menggunakan Algoritma Decision Tree (C4.5) Pada Indibiz Telkom Regional I

2026-06-11T21:27:27+07:00

This study analyzes the potential for customer churn among Indibiz Telkom Regional I business customers by implementing the Decision Tree C4.5 algorithm. The primary issue addressed is the high risk of losing business customers, which impacts revenue stability. This research develops an automated classification system utilizing customer behavior attributes such as payment status, total tickets, and total complaints. The research methodology includes data collection, preprocessing, calculating entropy and information gain, and constructing a decision tree. The results reveal that the "total ticket" attribute has the highest gain ratio, indicating that the frequency of service disruptions is the most dominant factor in triggering churn. Testing of the developed web-based system demonstrated an accuracy rate of 68% in classifying customers into churn and non-churn categories. The implementation of the C4.5 algorithm proves effective in mapping customer behavior patterns and serves as a decision support instrument for management to determine more targeted customer retention strategies.

Analisis Klaster Lagu Indonesia pada Platform Spotify Menggunakan Algoritma K-Means (Studi Kasus: Spotify)

2026-05-25T15:51:35+07:00

The growth of streaming platforms like Spotify has revolutionized music consumption in Indonesia, generating complex song performance data from Spotify Charts. This study aims to cluster Indonesian songs by popularity using the K-Means Clustering algorithm. The selected variables—peak, previous, streak, and streams—were chosen because they directly reflect song performance dynamics on the charts, encompassing best position, rank movement, longevity, and play volume, making them more representative of market resonance than audio features alone. A quantitative descriptive approach was applied to Indonesia Spotify Charts data spanning January 2020 to December 2024, with a purposive sample of 5,316 records post-cleaning. The optimal number of clusters was determined via the Elbow Method at k=3, with cluster quality evaluated using a Silhouette Score of 0.4968. Findings reveal three clusters: Cluster 0 (medium popularity, 3.8M average streams), Cluster 1 (low popularity, 569K average streams), and Cluster 2 (high popularity, 54M average streams). The study concludes that K-Means effectively identifies objective popularity patterns to support data-driven promotion strategies.

Implementasi Sistem QR Code Dalam Meningkatkan Tingkat Kepatuhan Pengisian Checklist Pemeriksaan Kendaraan Harian di PT XYZ: Studi Komparatif Sebelum dan Sesudah Penerapan

2026-05-25T18:13:22+07:00

Daily vehicle inspection is an important component of risk control in occupational safety and health (OSH) management systems. However, manual inspection processes often result in low compliance in checklist completion, reporting delays, and limited monitoring capabilities. This study aims to analyze the effect of implementing a QR Code system on the compliance level of daily vehicle inspection checklist completion at PT XYZ. The study employed a quasi-experimental design with a one-group pretest-posttest approach involving 35 operational vehicles. Data were collected from vehicle inspection checklists before and after the implementation of the QR Code system and analyzed using descriptive and inferential statistics through a paired sample t-test. The results showed that the compliance level increased from 72.2% before implementation to 90.2% after the implementation of the QR Code system. Statistical testing indicated a significant difference in compliance levels before and after implementation with t(34)=26.7 and p<0.001. An effect size value of 4.52 indicated a very large effect on compliance improvement. The implementation of the QR Code system proved effective in improving checklist completion, strengthening monitoring processes, and supporting the digitalization of occupational safety and health management systems in operational environments.

Analisis Komparatif Embedding Semantik Berbasis Large Language Model Pada Sistem Rekomendasi Buku Serendipitous di Perpustakaan Kampus

2026-05-25T19:03:38+07:00

The phenomenon of information overload in academic libraries often makes it difficult for users to discover relevant books, which may reduce reading interest. Conventional recommender systems are also prone to filter bubbles and tend to perform poorly under cold-start conditions. This study proposes a sequential recommendation system based on the Self-Attention Based Sequential Recommendation (SASRec) model integrated with five semantic embedding models, namely Word2Vec, BERT Multilingual, OpenAI text-embedding-3-small, Gemini-embedding-001, and Qwen3-Embedding-0.6B, to generate accurate and serendipitous recommendations. In addition, the Serendipity-Oriented Greedy (SOG) re-ranking algorithm is implemented to balance recommendation relevance and serendipity. The data set consists of 14,502 book records and 5,445 user interaction histories after the data cleaning process. Evaluation was conducted under three testing scenarios, namely the all-test set, warm test set, and cold test set, by comparing all model variants before and after the re-ranking process. The results show that the integration of Large Language Model (LLM)-based embeddings consistently improves performance compared to the standard SASRec model and traditional embeddings. Qwen3-Embedding-0.6B achieved the best performance, improving HitRate@10 by up to 282.9% and NDCG@10 by up to 387.8%, while maintaining semantic robustness in cold-start scenarios with an UnSerendipity@K score of 0.613. The implementation of SOG re-ranking reveals a direct trade-off between recommendation accuracy and diversity. Lightweight weighting provides the optimal balance, whereas overly aggressive weighting significantly reduces relevance. The main contribution of this study lies in integrating modern LLM embeddings into a sequential recommendation architecture to improve accuracy and cold-start robustness, while also evaluating the impact of serendipity-oriented re-ranking strategies on balancing recommendation relevance and diversity. Overall, this study demonstrates that modern LLM integration can produce a smarter, more adaptive, and more balanced library recommendation system in terms of both accuracy and serendipity.

Multilabel Aspect-Based Emotion Analysis Pada Ulasan Aplikasi IKD: Pengaruh Focal Loss dan Threshold Tuning Menggunakan Indobert

2026-06-14T21:30:11+07:00

User reviews of the Identitas Kependudukan Digital (IKD) application contain various emotions toward different service aspects. These reviews not only reflect the level of service satisfaction but also encompass user experiences, complaints, expectations, and public perceptions regarding the quality of the system. This study aims to develop a multi-label Aspect-Based Emotion Analysis (ABEA) model using an end-to-end IndoBERT architecture to identify user emotions across each service aspect of the IKD application. Additionally, it analyzes the impact of implementing Focal Loss and threshold tuning on classification performance under highly imbalanced label distributions. Data were collected from 13,197 user reviews on the Google Play Store spanning from June 2024 to November 2025 using web scraping methods, which were subsequently cleaned and filtered to yield 6,891 data entries. Service aspects were empirically identified using BERTopics. Labeling was conducted by three human annotators and two AI annotators, with the final labels determined through majority voting. The model was developed across 6 experimental scenarios varying in preprocessing, Focal Loss, threshold tuning, and data split ratios. Evaluation was performed using F1 Score Macro, F1 Score Micro, Precision, Recall, and Hamming Loss metrics. BERTopic achieved a Coherence Score of 0.6196 and a Topic Diversity of 0.92 with 5 representative aspects. The most optimal model was obtained using a configuration of Focal Loss, a threshold of 0.4, and a 60:20:20 split ratio, achieving an F1 Score Macro of 0.3916, a 24.1% increase from the baseline, alongside an F1 Score Micro of 0.9134 and a Recall of 0.9423. The selected model was successfully integrated into a web-based system using the Flask framework to visualize the classification results. Anger dominated the reviews concerning the Login & Akses Akun and Scan Barcode ke Dukcapil aspects, whereas the Dokumen & Layanan Digital aspect recorded the highest joy emotion. The combination of Focal Loss and threshold tuning proved effective in handling imbalanced label distributions in Indonesian multi-label ABEA classification.

Sistem Informasi Deteksi Kualitas Cabai Kopay Berbasis Image Processing dan Deep Learning dengan Python

2026-06-19T16:58:06+07:00

Kopay chili is a local agricultural commodity with high economic value. Manual quality assessment of Kopay chili is still subjective and time-consuming; therefore, a technology-based solution is required. This study develops an information system for detecting the quality of Kopay chili using image processing methods and deep learning based on Convolutional Neural Networks (CNN) implemented in Python. Key image features such as color, texture, and shape are analyzed to support the classification process. The dataset used consists of 2,000 images of Kopay chili categorized into good, medium, and poor quality classes. Experimental results show that the proposed system achieves an accuracy of 94.7% in classifying chili quality. This system provides an efficient and accurate solution for farmers and agribusiness stakeholders in evaluating agricultural product quality.

Comparative Analysis of Machine Learning Algorithms for Adolescent Depression Risk Classification Based on Digital Behavior Using SMOTE

2026-06-29T18:32:53+07:00

Adolescent depression is an increasingly alarming global mental health issue, particularly amid the rapid adoption of social media platforms. This study compares the performance of four machine learning algorithms Random Forest, XGBoost, Support Vector Machine (SVM), and Logistic Regression, in classifying adolescent depression risk using 12 digital behavioral and lifestyle features. The dataset comprises 1,200 records with severe class imbalance (97.4% vs 2.6%), addressed through Synthetic Minority Over-sampling Technique (SMOTE) with k=5. Eight experiments (four algorithms × two conditions) were conducted and evaluated using Precision, Recall, F1-Macro, AUC-ROC, and 5-fold cross-validation. Results demonstrate that SMOTE consistently enhanced all model performances, with an average F1-Macro improvement of 25.7%. XGBoost achieved perfect F1-Macro and AUC-ROC scores (1.0000) under both conditions, while SMOTE yielded the greatest gains for SVM (+67.9%) and Logistic Regression (+63.7%). Feature importance analysis identified sleep hours, stress level, anxiety level, and daily social media usage as dominant predictors, while academic performance contributed negligibly. These findings underscore the critical importance of imbalanced data handling and provide a methodological framework for developing digital behavior-based early depression detection systems.

Perancangan Antarmuka Q-Commerce Multi-Vertical Dengan Pendekatan Human-Centered Design Untuk Digitalisasi UMKM

2026-07-01T13:20:10+07:00

The growth of E-commerce in Indonesia has driven the emergence of Quick Commerce (Q-commerce), a location-based business model offering instant delivery services. However, existing platforms remain limited to specific product categories and are unable to accommodate urgent multi-vertical needs. This study aims to design a multi-vertical Q-commerce application interface using the Human-Centered Design (HCD) approach, consisting of three phases: inspiration, ideation, and implementation. The inspiration phase was conducted through comparative studies, interviews, and user questionnaires. The ideation phase produced functional requirements and a visual prototype using Figma. The implementation phase evaluated the prototype using the System Usability Scale (SUS) questionnaire. The results produced a SUS score of 75,92, falling under the Good and Acceptable category, suggesting that the designed interface is well-received and usable by the target users