Jurnal Politeknik Negeri Batam (PoliBatam)
Not a member yet
3001 research outputs found
Sort by
Detecting Fake Reviews in E-Commerce: A Case Study on Shopee Using Support Vector Machine and Random Forest
No full textThe increasing popularity of online shopping, particularly on platforms such as Shopee, has made product reviews a significant factor influencing consumer purchasing decisions. However, the presence of fake reviews generated by non-human agents undermines consumer trust and affects platform credibility. This study aims to detect fake reviews on Shopee by applying a text classification approach using Random Forest and Support Vector Machine (SVM) algorithms. A dataset consisting of 3,686 Shopee product reviews was collected and underwent preprocessing steps including data cleaning, normalization, tokenization, and TF-IDF weighting. Review labeling was performed automatically through the Latent Dirichlet Allocation (LDA) method, categorizing reviews into Original (OR) and Computer-Generated (CG). Model performance was evaluated using accuracy, precision, recall, and F1-score metrics. Experimental results show that the SVM algorithm achieved the highest accuracy at 88.84%, outperforming Random Forest which obtained 80.39%. These findings highlight the effectiveness of SVM in handling high-dimensional text data for fake review detection. The study contributes to the application of automated topic modeling (LDA) for labeling e-commerce reviews in the Indonesian context and opens opportunities for further enhancement using larger datasets and deep learning-based models to improve classification accuracy and scalability.The increasing popularity of online shopping, particularly on platforms such as Shopee, has made product reviews a significant factor influencing consumer purchasing decisions. However, the presence of fake reviews generated by non-human agents undermines consumer trust and affects platform credibility. This study aims to detect fake reviews on Shopee by applying a text classification approach using Random Forest and Support Vector Machine (SVM) algorithms. A dataset consisting of 3,686 Shopee product reviews was collected and underwent preprocessing steps including data cleaning, normalization, tokenization, and TF-IDF weighting. Review labeling was performed automatically through the Latent Dirichlet Allocation (LDA) method, categorizing reviews into Original (OR) and Computer-Generated (CG). Model performance was evaluated using accuracy, precision, recall, and F1-score metrics. Experimental results show that the SVM algorithm achieved the highest accuracy at 88.84%, outperforming Random Forest which obtained 80.39%. These findings highlight the effectiveness of SVM in handling high-dimensional text data for fake review detection. The study contributes to the application of automated topic modeling (LDA) for labeling e-commerce reviews in the Indonesian context and opens opportunities for further enhancement using larger datasets and deep learning-based models to improve classification accuracy and scalability
PERANCANGAN SISTEM INFORMASI AKUNTANSI PENERIMAAN DAN PENGELUARAN KAS PADA BUSINESS CENTER YAYASAN PERGURUAN DAARUSSALAAM JAGAKARSA
No full textBerdasarkan penelitian yang dilakukan pada Business Center (BC) Yayasan Perguruan Daarussalaam Jagakarsa, ditemukan bahwa sistem akuntansi yang diterapkan belum memanfaatkan teknologi digital dalam proses pencatatan transaksi penerimaan dan pengeluaran kas. Berdasarkan penelitian tersebut, peneliti merancang sebuah aplikasi berbasis Microsoft Excel yang disesuaikan dengan kebutuhan organisasi guna mendukung kelancaran aktivitas operasional. Penelitian ini menggunakan metode kualitatif dengan pendekatan studi kasus, serta memanfaatkan data primer dan sekunder dalam pengumpulan data. Pengembangan sistem dilakukan dengan metode System Development Life Cycle (SDLC) model waterfall. Hasil penelitian menunjukkan bahwa sistem akuntansi masih menggunakan buku besar folio yang rentan terhadap kesalahan pencatatan dan ketidaksesuaian data. Selain itu, pengelola BC merasa kesulitan dalam pembukuan akuntansi karena tidak memiliki latar belakang akuntansi dan tidak mendapatkan pelatihan di bidang tersebut. Oleh karena itu, peneliti memberikan solusi berupa rancangan sistem informasi akuntansi penerimaan dan pengeluaran kas berbasis Microsoft Excel yang telah terotomatisasi. Penerapan sistem ini mampu meningkatkan efisiensi, meminimalisasi kesalahan pencatatan, serta memperkuat akuntabilitas dan transparansi dalam pengelolaan kas sebagai bentuk pertanggungjawaban kepada pihak yayasan.Penelitian ini bertujuan untuk mengidentifikasi dan menganalisis sistem akuntansi penerimaan dan pengeluaran kas yang diterapkan pada Business Center (BC) Yayasan Perguruan Daarussalaam Jagakarsa, serta merancang sistem informasi akuntansi penerimaan dan pengeluaran kas berbasis Microsoft Excel yang sesuai dengan kebutuhan. Penelitian ini menggunakan metode kualitatif dengan pendekatan studi kasus. Proses perancangan sistem informasi akuntansi mengadopsi tiga tahapan dalam metode System Development Life Cycle (SDLC) model waterfall, yaitu analisis kebutuhan sistem, desain, dan code generation. Data yang digunakan meliputi data primer berupa wawancara dan observasi langsung, serta data sekunder dari dokumen terkait. Hasil penelitian menunjukkan bahwa sistem informasi akuntansi penerimaan dan pengeluaran kas pada BC Yayasan Perguruan Daarussalaam Jagakarsa masih dilakukan secara manual menggunakan buku besar folio bergaris, yang rentan terdapat kesalahan pencatatan dan ketidaksesuaian data. Selain itu, pengurus BC tidak memiliki latar belakang akuntansi dan tidak mendapatkan pelatihan di bidang tersebut. Oleh karena itu, peneliti memberikan rancangan sistem informasi akuntansi penerimaan dan pengeluaran kas berbasis Microsoft Excel yang terotomatisasi untuk mempermudah pencatatan dan pelaporan keuangan, serta meningkatkan efisiensi, akuntabilitas, dan transparansi dalam pengelolaan kas
Enhancing the Encryption Capabilities of the Generalization of the ElGamal Algorithm for Document Security
Get PDFThe development of cryptographic algorithms that are efficient in terms of computation and resource usage, in addition to maintaining the confidentiality, integrity, and authentication of information, is driven by the growing need for digital document security. The generalization of the ElGamal, an expansion of the traditional ElGamal algorithm with more adaptable encryption features, is one algorithm with a lot of promise in this situation. The implementation of the technique of splitting the plaintext into three-digit blocks to lower the complexity of encryption per character and the use of large prime numbers to increase the algorithm\u27s mathematical complexity are the two main ways that this study seeks to increase the algorithm\u27s efficiency and security. It is anticipated that this method will speed up computation time and simplify the encryption process per character without compromising security. The overall findings demonstrate that, without compromising security, this method dramatically cuts down on computation time and ciphertext enlargement. Therefore, in the age of digital transformation, the findings of this study aid in the creation of contemporary cryptographic algorithms that are more flexible and effective and serve as a strategic guide when creating a strong digital data security system
Comparative Analysis of ResNet50V2, ResNet152V2, and MobileNetV2 Architectures in Monkeypox Classification
Get PDFConvolutional Neural Networks (CNN) are recognized for their high accuracy in image classification, but large-scale datasets and significant computer resources are needed to train them from scratch, though. Transfer learning offers a practical solution by leveraging pre-trained models to accelerate training even when data is limited. Although CNNs have been widely applied to skin disease classification, specific evaluations of architectures such as ResNet50V2, ResNet152V2, and MobileNetV2 for monkeypox image classification remain scarce. Therefore, this study aims to comprehensively compare the effectiveness and trade-offs of these architectures in detecting monkeypox through transfer learning. The evaluation focuses on balancing accuracy and computational efficiency across stages, including data collection, preprocessing, model design, training, and testing. The dataset, obtained from Kaggle, consists of 2,310 images across four classes: monkeypox, chickenpox, measles, and normal. Transfer learning was implemented using fine-tuned weights from ImageNet. According to the results, ResNet152V2 needed the most training time but had the lowest loss and the greatest validation accuracy (98.28%). ResNet50V2 maintained a good compromise between accuracy (97.84%) and training efficiency, while MobileNetV2 yielded the best overall classification metrics (97.86% for accuracy, precision, recall, and F1-score), indicating strong generalization. These findings highlight the distinct strengths of each model, offering insights into architecture selection based on specific operational constraints and goals
Detection of Qur’anic Ikhfa Patterns in Digital Images Using Binary Similarity Distance Measures (BSDM) with 3W-Jaccard Formula
Get PDFRecitation rules in the Qur\u27anic script form various visual patterns. One of the selected rules for this study is the Ikhfa pattern. Ikhfa is a recitation rule pronounced subtly when the nun sukun (نْ) or tanwin (ـَــًـ, ـِــٍـ, ـُــٌـ) is followed by one of 15 specific letters, namely: ta’ (ت), tsa’ (ث), jim (ج), dal (د), dzal (ذ), za’ (ز), sin (س), syin (ش), shad (ص), dhad (ض), tha’ (ط), zha’ (ظ), fa’ (ف), qaf (ق), and kaf (ك). In this study, the primary challenge is the difficulty of automatically detecting the Ikhfa pattern in both digital and printed Qur\u27anic texts. This challenge arises from the subtlety of the recitation rule, which makes it difficult to distinguish from other recitation patterns. To address this, the Ikhfa pattern is detected using image processing techniques, and pattern classification is performed using the Binary Similarity and Distance Measures (BSDM) method. The results indicate that the pattern detection system, employing BSDM with the 3W-Jaccard formula, achieved a detection rate of 83.84%. This suggests that the 3W-Jaccard formula is an effective approach for detecting similar recitation patterns. One advantage of the 3W-Jaccard formula is its ability to recognize patterns with a relatively small amount of reference data, making it highly suitable for implementation in the detection system
Comparison of FMADM TOPSIS and FMADM WP in Determining Recipients of the Family Hope Program (PKH) Assistance
Get PDFFuzzy Multi-Attribute Decision Making (FMADM) TOPSIS and WP methods are frequently employed to identify potential recipients of government assistance. The Family Hope Program (PKH) is a government social assistance program designed to improve the welfare of underprivileged individuals. However, the process of distributing this assistance often faces obstacles in the form of inaccuracy in determining recipients. This study compares FMADM TOPSIS and WP to evaluate their effectiveness in objectively determining potential PKH recipients. The criteria for potential PKH recipients are eleven criteria obtained from the social service based on government regulations and PKH assistants. Meanwhile, the alternatives for this study are fifty samples of family data for potential PKH recipients. This study employs a sensitivity test method to assess the accuracy of the results obtained from each method. The results of the study show that FMADM TOPSIS produces a higher level of accuracy of 94% compared to FMADM WP. This study is expected to be able to contribute to choosing the right decision-making method to determine potential recipients of social assistance.Fuzzy Multi-Attribute Decision Making (FMADM) TOPSIS and WP methods are frequently employed to identify potential recipients of government assistance. The Family Hope Program (PKH) is a government social assistance program designed to improve the welfare of underprivileged individuals. However, the process of distributing this assistance often faces obstacles in the form of inaccuracy in determining recipients. This study compares FMADM TOPSIS and WP to evaluate their effectiveness in objectively determining potential PKH recipients. The criteria for potential PKH recipients are eleven criteria obtained from the social service based on government regulations and PKH assistants. Meanwhile, the alternatives for this study are fifty samples of family data for potential PKH recipients. This study employs a sensitivity test method to assess the accuracy of the results obtained from each method. The results of the study show that FMADM TOPSIS produces a higher level of accuracy of 94% compared to FMADM WP. This study is expected to be able to contribute to choosing the right decision-making method to determine potential recipients of social assistance
UI/UX Optimization of GOBIS Suroboyo Application with User Centered Design Approach and Short User Experience Questionnaire
Get PDFGOBIS Suroboyo is a mobile application designed to assist Suroboyo Bus passengers in accessing route information, schedules, and general bus details. Despite its potential, the application has lacked systematic user experience evaluation, resulting in usability issues that require improvement. This study aims to optimize the user interface (UI) and user experience (UX) of the GOBIS Suroboyo application using the User-Centered Design (UCD) approach. The research was conducted through four main stages: analysis of the existing application, identification of user needs, redesign of the interface, and evaluation of the resulting prototype. The usability evaluation was performed using the Short User Experience Questionnaire (UEQ-S), which assessed both hedonic and pragmatic qualities. The results showed mean scores of 1.69 for hedonic quality and 1.425 for pragmatic quality, which fall into the "Excellent" and "Above Average" categories, respectively, based on the benchmark scale. These results indicate that the redesigned prototype is engaging, motivating, efficient, and user-friendly. This study concludes that the UCD approach, with active user involvement, is effective in enhancing the user experience of mobile applications
Design of an IoT-Based Air Quality System with Web Integration in a Palm Oil Mill Environment
Get PDFThis research aims to design and implement an Internet of Things (IoT)-based air quality monitoring system with real-time data integration to web applications in the PT Gemareksa Mekarsari palm oil mill environment. This system utilizes NodeMCU ESP32 as the main microcontroller connected with MQ-2 and MQ-135 sensors to detect CO, CH₄, and NH₃ gases. Data is sent in real-time to the ThingSpeak platform and displayed through a responsive web dashboard. Testing was conducted over two days at three different location points (open area, fruit processing, and factory office) with a total of 45 measurements. Results showed that the system was able to transmit data with a delivery accuracy rate of 86.67%, with most data received without delay.. The detected gas concentrations were within safe limits, although mild fluctuations occurred, especially in the fruit processing area. The system also showed stable performance in displaying data on mobile and desktop devices. Thus, this system can be an effective solution for automatic and real-time industrial air monitoring, and support efforts to mitigate health risks due to air pollution.This research aims to design and implement an Internet of Things (IoT)-based air quality monitoring system with real-time data integration to web applications in the PT Gemareksa Mekarsari palm oil mill environment. This system utilizes NodeMCU ESP32 as the main microcontroller connected with MQ-2 and MQ-135 sensors to detect CO, CH₄, and NH₃ gases. Data is sent in real-time to the ThingSpeak platform and displayed through a responsive web dashboard. Testing was conducted over two days at three different location points (open area, fruit processing, and factory office) with a total of 45 measurements. Results showed that the system was able to transmit data with a delivery accuracy rate of 86.67%, with most data received without delay.. The detected gas concentrations were within safe limits, although mild fluctuations occurred, especially in the fruit processing area. The system also showed stable performance in displaying data on mobile and desktop devices. Thus, this system can be an effective solution for automatic and real-time industrial air monitoring, and support efforts to mitigate health risks due to air pollution
Application of Linear Discriminant Analysis Method With Gray Level Cooccurrence Matrix Method for Classification of Lung Disease Diagnosis Based on X-Ray Results
Get PDFThis study aims to classify lung diseases from X-ray images using a combination of Gray Level Cooccurrence Matrix (GLCM) and Linear Discriminant Analysis (LDA) methods. GLCM was used to extract texture features such as contrast, correlation, energy, and homogeneity from 300 lung X-ray images representing four categories: Normal, Pneumonia, Tuberculosis, and Bronchitis. The LDA method was then applied for classification based on these features. The results showed that Tuberculosis had the highest classification accuracy at 80%, while the overall model accuracy was 61.67%. Evaluation using precision, recall, F1-score, and confusion matrix confirmed that the GLCM and LDA combination performed best in identifying tuberculosis cases. However, overlapping features between Normal, Bronchitis, and Pneumonia classes reduced the classification performance. These findings suggest that the proposed method provides promising results and could be improved further with advanced feature extraction or classification techniques
Evaluating the Performance of an LBS-Based Waste Reporting Application for Digital Waste Management
Get PDFThe escalating volume of urban waste in Indonesia presents a serious challenge, exacerbated by conventional reporting mechanisms that are slow and inefficient. This study aims to develop and evaluate E-Trash, a Location-Based Services (LBS) application designed to accelerate the workflow of participatory waste reporting, handling, and monitoring in Makassar City. The novelty of this research lies in the synergistic integration of citizen reporting, real-time bidirectional notifications between reporters and field officers, and a spatial monitoring dashboard for policymakers, validated through direct, real-world implementation. The research methodology employs a software engineering approach utilizing a prototype model. System validation was conducted in three stages: black-box testing on 24 core features, performance testing under various bandwidth conditions, and a two-week field trial involving community members and sanitation personnel in two sub-districts. The findings robustly conclude that the E-Trash application effectively leverages a digital, Location-Based Services (LBS) approach to significantly enhance citizen participation in waste reporting and improve the response efficiency of sanitation personnel. The system demonstrated optimal functionality across diverse network conditions and device types, with stable response times and a high data transmission success rate affirming its reliability. Field implementation notably yielded a reduction in illegal waste accumulation and an increase in overall handling efficiency, primarily facilitated by the bidirectional notification system between citizens and sanitation teams. Consequently, E-Trash emerges as a highly viable candidate for replication in other urban settings, serving as a robust, community-participation-centric smart solution for sustainable urban sanitation management