Jurnal Politeknik Negeri Batam (PoliBatam)
Not a member yet
3001 research outputs found
Sort by
PENDAMPINGAN TEKNIS PENCATATAN KEUANGAN DAN LEGALITAS USAHA BAGI USAHA MIKRO PEMULA DI KEPULAUAN RIAU
Get PDFMicro, Small and Medium Enterprises (MSMEs) have an essential role in the Indonesian economy, but still face significant challenges in financial management and business legality, especially in the Riau Islands with limited access to training. This community service program aims to provide technical assistance in financial recording and business legality for budding micro businesses in the Riau Islands to create independent and sustainable entrepreneurs. Through the Project-Based Learning (PBL) approach, Batam State Polytechnic students accompany micro business actors in simple financial recording and processing legalities such as Business Identification Number (NIB), Taxpayer Identification Number (NPWP), and halal certification. As a result, 80% of participants succeeded in implementing daily transaction recording and 70% in compiling simple financial reports. In comparison, 75% of participants managed business legality through the Online Single Submission (OSS) platform, which increased access to formal financing. This program also changes the participants\u27 mindset towards more professional business management, although further assistance is needed to ensure consistent implementation. This program has the potential to be replicated in other regions to support more competitive and sustainable empowerment of MSMEs, as well as contribute to local economic growth.Usaha Mikro, Kecil, dan Menengah (UMKM) memiliki peran penting dalam perekonomian Indonesia, namun masih menghadapi tantangan besar dalam pengelolaan keuangan dan legalitas usaha, terutama di Kepulauan Riau dengan keterbatasan akses terhadap pelatihan. Program pengabdian masyarakat ini bertujuan memberikan pendampingan teknis pada pencatatan keuangan dan legalitas usaha bagi usaha mikro pemula di Kepulauan Riau, guna menciptakan wirausaha mandiri dan berkelanjutan. Melalui pendekatan Project-Based Learning (PBL), mahasiswa Politeknik Negeri Batam mendampingi pelaku usaha mikro dalam pencatatan keuangan sederhana dan pengurusan legalitas seperti Nomor Induk Berusaha (NIB), Nomor Pokok Wajib Pajak (NPWP), dan sertifikasi halal. Hasilnya, 80% peserta berhasil menerapkan pencatatan transaksi harian dan 70% menyusun laporan keuangan sederhana, sementara 75% peserta mengurus legalitas usaha melalui platform Online Single Submission (OSS), yang meningkatkan akses pembiayaan formal. Program ini juga mengubah pola pikir peserta terhadap pengelolaan usaha yang lebih profesional, meskipun pendampingan lanjutan diperlukan untuk memastikan penerapan konsisten. Program ini berpotensi direplikasi di wilayah lain untuk mendukung pemberdayaan UMKM yang lebih kompetitif dan berkelanjutan, serta berkontribusi pada pertumbuhan ekonomi lokal
Peramalan Jumlah Permintaan Crumb Rubber SIR 20 Menggunakan Metode Backpropagation Pada PT XYZ
No full textPT XYZ merupakan industri manufaktur yang memproduksi crumb rubber dengan jenis SIR 20. Tahun 2021, perbandingan jumlah persediaan bahan baku dengan hasil produksi memiliki tingkat eror MAPE lebih dari 50% yang mengindikasikan bahwa model yang digunakan tergolong buruk. Selisih antara jumlah persediaan bahan baku dan jumlah produksi berpotensi meningkatkan biaya produksi. Jumlah persediaan bahan baku di atas titik optimal dapat menyebabkan meningkatnya biaya penyimpanan, sedangan jumlah persediaan di bawah titik optimal berpotensi menambah biaya produksi akibat biaya kekurangan. Tujuan penelitian ini adalah memilih model arsitektur Backpropagation dengan nilai MAPE terendah yang selanjutnya digunakan memprediksi jumlah permintaan produk SIR 20 pada tahun 2023. Arsitektur Backpropagation yang digunakan pada penelitian ini adalah (3-9-1-1), (3-10-1-1), (3-11-1-1), (3-11-1-1), dan (3-12-1-1). Variabel input dalam penelitian ini adalah tahun produksi, jumlah hari, dan harga SIR 20. Model Backpropagation (3-11-1-1) merupakan model yang memberikan hasil prediksi dengan eror terendah, yaitu 14,5%. Model tersebut memberi hasil penuruan nilai MAPE sebesar 36,5% dan peningkatan pendapatan berdasarkan selisih antara pembelian bahan baku dan penjual produk sebesar Rp 62.295.269.963,13
Smart System for Early Diagnosis of Gastroesophageal Reflux Disease (GERD)
No full textGastroesophageal Reflux Disease (GERD) occurs when gastric contents reflux into the esophagus, yet early diagnosis remains limited due to the lack of accessible screening tools. This limitation contributes to reduced public awareness and delays in seeking appropriate medical evaluation. To address this problem, this study aims to develop an intelligent system capable of supporting early GERD diagnosis. The proposed system evaluates stomach acid levels through saliva pH measurement and incorporates symptom assessment using the GERD-Q questionnaire, which is widely adopted by internist physicians as a clinical screening instrument. Additional variables—including lifestyle factors, age, height, and weight—are integrated into a logistic regression model to estimate the probability of GERD. The pH sensor demonstrates an accuracy of approximately 99,25%. Future studies will focus on validating the sensor data against patient medical records and comparing the system’s diagnostic performance with standard clinical examinations conducted by healthcare professionals
Comparative Study of Manual and Generated Data Transfer Object Implementation Performance
No full textThe Data Transfer Object (DTO) is a fundamental component in Flutter application development, particularly in managing data serialization and deserialization. This study compares two DTO implementation methods—manual and generated—focusing on execution speed and memory efficiency. Testing was conducted across three levels of data complexity (Small, Medium, and Large) over 100 iterations using Flutter DevTools. The findings reveal that the generated approach (utilizing libraries such as json_serializable) consistently outperforms the manual approach. Specifically, it achieves a 1:1.147 ratio in parsing speed and a 1:1.42 ratio in memory efficiency compared to manual DTOs. Although the manual method provides greater flexibility for implementing conditional parsing logic, it tends to be more error-prone and less efficient when handling large datasets. In contrast, the generated approach offers faster performance, better scalability, and reduced human error potential, making it the preferred option for projects demanding technical efficiency and rapid development cycles. Consequently, this study recommends adopting generated DTOs for applications dealing with large-scale and complex data, while reserving manual DTOs for cases requiring highly dynamic or conditional data parsing
Classification For Determining Nutritional Status of Toddlers Using Random Forest Method at Tanah Pasir Primary Health Centre, North Aceh
No full textThe nutritional status of toddlers is a fundamental factor in supporting their growth and development, particularly during the golden period of 0–5 years of age. Malnutrition in toddlers can have detrimental effects on physical growth, cognitive development, and immune function. In Indonesia, child malnutrition remains a significant public health challenge, particularly in rural areas, necessitating improved nutritional surveillance systems at primary health centers. The manual assessment of nutritional status at community health centers (Puskesmas) often poses challenges in promptly identifying toddlers with undernutrition or severe malnutrition. This study aims to develop a toddler nutritional status classification system based on the Random Forest method to assist healthcare workers in determining nutritional status quickly and accurately. This study utilized a dataset of 2,612 toddler anthropometric records collected from Tanah Pasir Community Health Center, North Aceh, between November 2024 and January 2025. The dataset was split into training (2,090 records, 80%) and testing (522 records, 20%) sets using stratified random sampling. Key variables included age (0-60 months), body weight (kg), and body height (cm). Nutritional status categories were determined based on WHO Child Growth Standards using the weight-for-age (W/A), height-for-age (H/A), and weight-for-height (W/H) indices. The Random Forest method was chosen due to its ability to construct multiple decision trees through ensemble learning, resulting in more accurate predictions and better resistance to overfitting. The model was implemented with 100 trees and evaluated using standard classification metrics. The experimental results demonstrated that the system achieved strong classification performance, with an accuracy of 93%, precision of 95%, recall of 98%, and an F1-score of 96%. The high recall value is particularly significant in healthcare applications, ensuring minimal false negatives in detecting malnourished toddlers. The developed system facilitates healthcare workers in efficiently and systematically monitoring toddlers\u27 nutritional status with consistent classification standards. Therefore, this system is expected to serve as a decision-support tool to improve community nutritional status at the community health center level, enabling early intervention for at-risk children
Comparison of Support Vector Machine (SVM) and Random Forest Algorithms in the Analysis of SOcial Media X User Sentiment Towards the TNI Bill
No full textThe rapid advancement of information technology has enabled the public to openly express their views through social media, including on strategic national issues such as the Draft Law on the Indonesian National Armed Forces (RUU TNI). This study aims to map public sentiment toward the RUU TNI and to compare the effectiveness of two popular sentiment analysis algorithms, Support Vector Machine (SVM) and Random Forest (RF). A total of 525 relevant tweets collected between February and May 2025 were analyzed and classified into three sentiment categories: positive, negative, and neutral. The results reveal that neutral opinions dominate at 81.4%, followed by negative sentiments at 11.1% and positive sentiments at 7.4%. The performance comparison shows that SVM achieved an accuracy of 92%, outperforming RF which obtained 91%. These findings highlight that strategic defense issues tend to generate predominantly informative public opinions, while critical voices show an increasing trend as the discourse evolves. The novelty of this study lies in the application of three-class sentiment classification and the comparative evaluation of SVM and RF within the domain of defense policy. This research contributes to the academic discourse by extending sentiment analysis beyond electoral and marketing topics, while also providing practical insights for policymakers in understanding and responding to public aspirations more effectively.The rapid advancement of information technology has enabled the public to openly express their views through social media, including on strategic national issues such as the Draft Law on the Indonesian National Armed Forces (RUU TNI). This study aims to map public sentiment toward the RUU TNI and to compare the effectiveness of two popular sentiment analysis algorithms, Support Vector Machine (SVM) and Random Forest (RF). A total of 525 relevant tweets collected between February and May 2025 were analyzed and classified into three sentiment categories: positive, negative, and neutral. The results reveal that neutral opinions dominate at 81.4%, followed by negative sentiments at 11.1% and positive sentiments at 7.4%. The performance comparison shows that SVM achieved an accuracy of 92%, outperforming RF which obtained 91%. These findings highlight that strategic defense issues tend to generate predominantly informative public opinions, while critical voices show an increasing trend as the discourse evolves. The novelty of this study lies in the application of three-class sentiment classification and the comparative evaluation of SVM and RF within the domain of defense policy. This research contributes to the academic discourse by extending sentiment analysis beyond electoral and marketing topics, while also providing practical insights for policymakers in understanding and responding to public aspirations more effectively
Real-Time Arrow Detection and Scoring on Archery Targets Using YOLOv8 with Euclidean Distance-Based Zone Estimation
No full textThe current study aims to create an automated scoring system for archery target board using computer vision technologies. As archery has develop from a traditional practice to a competitive activity, the scoring procedures have become a crucial element. While the current manual scoring procedures are fallible and can be challenging for organizers. This study offers a solution to this issue by using YOLO v8 (You Only Look Once) architecture for real- time arrow recognition and scoring. The development process consists of dataset collecting, picture pre-processing, model training and implementation using 2 photos of the target boards with arrows. The computer processes the scores by calculating the distance from the center of the arrow to the selected scoring zones using Euclidean distance. System testing established a baseline accuracy of 67%. While users noted the system\u27s processing efficiency (speed), this accuracy level highlights significant room for improvement. The results demonstrate the potential for applying computer vision to automate the archery scoring system, while simultaneously emphasizing the critical need for advanced model performance enhancements. This study serves as a preliminary step in exploring automated sport technology, expected to contribute to future refinements of the archery scoring system.The current study aims to create an automated scoring system for archery target board using computer vision technologies. As archery has develop from a traditional practice to a competitive activity, the scoring procedures have become a crucial element. While the current manual scoring procedures are fallible and can be challenging for organizers. This study offers a solution to this issue by using YOLO v8 (You Only Look Once) architecture for real- time arrow recognition and scoring. The development process consists of dataset collecting, picture pre-processing, model training and implementation using 2 photos of the target boards with arrows. The computer processes the scores by calculating the distance from the center of the arrow to the selected scoring zones using Euclidean distance. System testing established a baseline accuracy of 67%. While users noted the system\u27s processing efficiency (speed), this accuracy level highlights significant room for improvement. The results demonstrate the potential for applying computer vision to automate the archery scoring system, while simultaneously emphasizing the critical need for advanced model performance enhancements. This study serves as a preliminary step in exploring automated sport technology, expected to contribute to future refinements of the archery scoring system
Development of an IoT-Based Smart Cane with Non-Invasive Health Monitoring for Elderly Care in Batam
No full textThe rapid growth of the elderly population requires assistive technologies that support mobility, health, and safety. This study presents the development of an IoT-based smart cane designed to enhance elderly independence and health monitoring in Batam, Indonesia. The prototype integrates non-invasive health sensors (MAX30102 for heart rate and SpO₂, MLX90614 for temperature, and a non-invasive glucose sensor), a GPS module, a mini-CCTV with two-way audio, and a solar-powered energy system, all controlled by an ESP32 microcontroller connected to the Blynk IoT platform. Ergonomic design was guided by anthropometric data of Indonesian elderly to ensure user comfort and usability. Experimental results demonstrated stable performance of the integrated modules. Heart rate values ranged from 86–103 BPM (mean 89.5 ± 6.2 BPM), blood glucose estimations from 110–112 mg/dL (mean 111 ± 0.9 mg/dL), and body temperature from 36.9–37.1 °C (mean 37.0 ± 0.1 °C), all of which aligned closely with clinical references. Oxygen saturation readings, however, averaged 89 ± 0.8%, slightly below the clinical norm (≥95%), highlighting the need for sensor calibration. Dynamic testing of the GPS module across a 500-meter route achieved positional accuracy within 3–5 meters, while the CCTV system successfully streamed live video but was dependent on WiFi stability.The novelty of this research lies in the unique combination of locally adapted ergonomic design, multi-sensor non-invasive health monitoring, two-way visual and audio communication, GPS tracking, and renewable energy integration within a single portable device. These contributions not only enrich IoT-based healthcare research but also provide practical solutions tailored to elderly care in Indonesia. Future work will focus on clinical-grade validation of sensors, extended field trials, and the integration of predictive analytics using Machine Learning and Fuzzy Logic.The rapid growth of the elderly population requires assistive technologies that support mobility, health, and safety. This study presents the development of an IoT-based smart cane designed to enhance elderly independence and health monitoring in Batam, Indonesia. The prototype integrates non-invasive health sensors (MAX30102 for heart rate and SpO₂, MLX90614 for temperature, and a non-invasive glucose sensor), a GPS module, a mini-CCTV with two-way audio, and a solar-powered energy system, all controlled by an ESP32 microcontroller connected to the Blynk IoT platform. Ergonomic design was guided by anthropometric data of Indonesian elderly to ensure user comfort and usability. Experimental results demonstrated stable performance of the integrated modules. Heart rate values ranged from 86–103 BPM (mean 89.5 ± 6.2 BPM), blood glucose estimations from 110–112 mg/dL (mean 111 ± 0.9 mg/dL), and body temperature from 36.9–37.1 °C (mean 37.0 ± 0.1 °C), all of which aligned closely with clinical references. Oxygen saturation readings, however, averaged 89 ± 0.8%, slightly below the clinical norm (≥95%), highlighting the need for sensor calibration. Dynamic testing of the GPS module across a 500-meter route achieved positional accuracy within 3–5 meters, while the CCTV system successfully streamed live video but was dependent on WiFi stability.The novelty of this research lies in the unique combination of locally adapted ergonomic design, multi-sensor non-invasive health monitoring, two-way visual and audio communication, GPS tracking, and renewable energy integration within a single portable device. These contributions not only enrich IoT-based healthcare research but also provide practical solutions tailored to elderly care in Indonesia. Future work will focus on clinical-grade validation of sensors, extended field trials, and the integration of predictive analytics using Machine Learning and Fuzzy Logic
Design and Implementation of a Backend System and DevOps Workflow for Interactive Learning Applications
No full textEnglish language learning in Indonesia faces significant challenges, including limited vocabulary retention, poor pronunciation, and passive learning methods. The EngVenture application was developed to address these issues by integrating gamification principles with interactive English learning environments. This study aims to design and implement a backend system and DevOps workflow that ensure optimal performance, security, and stability for gamification-based learning applications. The Rapid Application Development (RAD) method was employed, comprising requirements planning, user design, construction, and cutover phases. System requirements were identified through a validated questionnaire (Cronbach\u27s α = 0.89) distributed to 101 respondents from diverse backgrounds. Results indicated that users prioritized data security (90.1%), system speed (91.1%), and secure authentication (69.3%) as critical factors. Based on these findings, a RESTful API-based backend was designed and integrated with Docker, Jenkins, and Nginx, incorporating security features such as JWT authentication, API key validation, and SSL/TLS encryption. Quantitative evaluation over a 20-day period demonstrated significant improvements: 85% faster deployment time (6.23→1.48 minutes), 43.4% reduction in error rate (211→138 errors), 95.7% build success rate, stable API response time (~160ms) under load testing with 1,000 concurrent requests, and near-zero downtime (<5 minutes). This research demonstrates that the integration of structured backend architecture and automated DevOps practices significantly enhances system reliability, deployment efficiency, and user satisfaction in educational technology applications such as EngVenture
Comparison of Multiple Linear Regression and Random Forest Methods for Predicting National Rice Production in Indonesia
No full textRice is a strategic commodity that plays an important role in maintaining national food security. However, rice production in Indonesia still fluctuates due to variations in harvest area, productivity, climate conditions, and differences in regional characteristics. This condition demands a predictive model capable of providing more accurate production estimates to support food policy planning. This research aims to predict national rice production by comparing two methods: Multiple Linear Regression and Random Forest Regression, using data from the Central Bureau of Statistics (BPS) and Nasa Power for the period 2018–2024. The analysis stages include data preprocessing, data exploration, categorical variable transformation, splitting data into training and testing sets, model training, and evaluation using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and the coefficient of determination (R²). The research results show that harvested area is the most dominant factor influencing rice production, followed by productivity, year, and province. Based on the evaluation results, Random Forest provided the best performance with an MAE value of 40,599.94, an RMSE of 77,153.07, and an R² of 0.9991. The low error value and the proximity of the prediction to the actual data indicate that this model is better at capturing non-linear patterns and inter-regional variations compared to Multiple Linear Regression. Overall, Random Forest can be an effective method for predicting national rice production and can be further developed in subsequent research by incorporating climate variables or other external factors