Jurnal Sistemasi (OJS FTIK - UNISI, Fakultas Teknik dan Ilmu Komputer Universitas Islam Indragiri)
Not a member yet
946 research outputs found
Sort by
Comparative Analysis of Deep Learning Architectures for Indonesian Spice Image Classification
Full text linkSpices are an important commodity in Indonesia; however, visual identification remains challenging due to the similarity in appearance among different types of spices. This study develops an image classification system for Indonesian spices using transfer learning by comparing four Convolutional Neural Network (CNN) architectures: VGG16, ResNet50, EfficientNetB0, and MobileNetV2. The Indonesian Spices dataset consists of 31 classes with a total of 6,510 images, which were stratified and divided into training, validation, and testing sets. The training process was conducted in two stages: head-layer training and fine-tuning, with the application of regularization techniques such as dropout, batch normalization, and L2 regularization. The results show that ResNet50 achieved the best performance with a test accuracy of 95.80%, followed by VGG16 with 95.70%. EfficientNetB0 provided an optimal balance between accuracy (94.17%) and the fastest inference time (5.51 ms), while MobileNetV2 achieved an inference time of 6.07 ms with an accuracy of 92.63%, making it suitable for mobile devices. This study demonstrates the effectiveness of transfer learning for Indonesian spice image classification
Development of an Internet of Things (IoT)-based Air Quality Monitoring System in the Environment
Full text linkThis study developed an Internet of Things (IoT)-based air quality monitoring system to measure temperature, humidity, and carbon dioxide (CO₂) levels in real time. The study employed a prototyping method consisting of problem identification, requirements analysis, system design and development, testing, and result analysis. The system utilizes a NodeMCU V3 as the microcontroller, an MQ-135 sensor for gas detection, a DHT11 sensor for temperature and humidity measurement, and an OLED LCD for local display. Measurement data are transmitted and stored on the ThingSpeak platform and can be accessed through an Android application. Testing was conducted under three conditions: a normal environment, a closed room without ventilation, and a polluted condition with cigarette smoke exposure. The results show that the system is able to responsively detect changes in air quality, with CO₂ levels recorded at 76 ppm under normal conditions, 183 ppm in the closed room, and 729 ppm in the polluted condition. The system operates stably and provides real-time data visualization, making it suitable for low-cost implementation in household environments and small communities
Analysis of User Acceptance of the Access by KAI Online Ticket Booking Application using TAM
Full text linkInformation technology (IT) has transformed ticketing systems in public transportation and other service industries. The Access by KAI online application, developed by PT Kereta Api Indonesia (Persero), facilitates train ticket purchases. This application features electronic boarding, travel itinerary information, and ticket booking services, thereby streamlining the user’s travel experience. However, users still raise concerns regarding the usability of the application. This study employs the Technology Acceptance Model (TAM) framework to assess user acceptance of the Access by KAI application. This quantitative study collected data from Access by KAI users through a questionnaire survey. Actual System Use (ASU) serves as the dependent variable, while Perceived Ease of Use (PEOU), Perceived Usefulness (PU), Attitude Toward Use (ATU), and Behavioral Intention (BI) function as independent and mediating variables. The data were analyzed using Structural Equation Modeling–Partial Least Squares (SEM-PLS) with SmartPLS software. The results show that four out of five hypotheses are supported, while one is not. The analysis indicates that Perceived Usefulness and Perceived Ease of Use significantly influence Attitude Toward Use and Perceived Usefulness. Another important finding is that Behavioral Intention and Attitude Toward Use have a strong influence on Actual System Use. However, the model does not demonstrate a significant effect for one of the hypothesized relationships. Overall, the findings suggest that users’ attitudes and intentions toward using the Access by KAI application are influenced by their perceptions of its ease of use and the benefits it provides. This study is expected to provide PT Kereta Api Indonesia (Persero) with evaluative insights to enhance service quality, improve user acceptance, and increase the usability of the Access by KAI application
Aspect-based Sentiment Analysis of Public Opinions on Integrated Islamic Schools using Lexicon based and Machine Learning Approaches
Full text linkThis study aims to examine public perceptions of Integrated Islamic Schools through aspect-based sentiment analysis by integrating Latent Dirichlet Allocation, Lexicon-Based approach, and Deep Neural Networks. LDA is employed to extract topic structures that represent the semantic context of public reviews, Lexicon Based method is used for sentiment analysis, while DNN infers sentiment orientation based on the extracted representations. This approach seeks to combine the strengths of probabilistic topic modeling and deep learning to obtain a more comprehensive understanding of public opinion. The analysis was conducted on a collection of 2,280 online reviews, which after preprocessing resulted in 1,438 reviews processed using the LDA–DNN combination. The results demonstrate that this approach is capable of identify in opinion dimensions in a more contextual manner and enhancing the interpretability of the analysis outcomes. Empirical evaluation shows that the proposed model achieved an accuracy of 63.89% for aspect classification and 93.06% for sentiment classification, outperforming the K-Means–LSA and K-Means–PCA approaches, which achieved 45.14% and 31.94% accuracy for aspect classification and 92.36% accuracy for sentiment classification, respectively. These findings confirm the superiority of probabilistic topic modeling in capturing complex semantic relationships and provide a methodological contribution to the development of sentiment analysis in the context of integrated Islamic education
Comparison of Filter and Wrapper Feature Selection Methods for Heart Disease Risk Classification using K-Nearest Neighbors (k-NN)
Full text linkFeature selection plays a crucial role in improving the effectiveness of medical classification models. This study compares two feature selection approaches—filter and wrapper methods—in developing a k-Nearest Neighbors (k-NN) model for heart disease risk classification. The dataset consists of patients’ demographic data, lifestyle factors, and clinical indicators. In this study, the filter method was applied by considering data types: Pearson Correlation was used for numerical features, while the Chi-Square test was applied to categorical features. The selected features from both techniques were then combined, reducing the initial 20 features to four key variables considered most relevant for heart disease risk classification: BMI, homocysteine level, blood pressure, and stress level. This approach achieved high computational efficiency; however, it resulted in only a modest accuracy improvement (76.8%) and a low recall for the minority class (0.07). In contrast, the wrapper method using Sequential Forward Selection (SFS) produced a more informative subset of 11 features, achieving higher accuracy (80.00%) and a ROC-AUC of 0.657, indicating better discrimination capability for the minority class. These findings suggest that while the filter method excels in simplicity and computational efficiency, the wrapper method is more effective in improving classification performance. This study provides empirical insights into selecting appropriate feature selection strategies based on analytical objectives, particularly for clinical decision support systems
Evaluation of User Satisfaction in Web-based Library Information Systems: A Systematic Literature Review
Full text linkThe transformation of library management today is highly influenced by the acceleration of information and communication technology (ICT), particularly through the adoption of web-based information systems. While these systems can optimize productivity and service accessibility, their effectiveness ultimately depends on the level of user satisfaction. This study evaluates various user satisfaction assessment methodologies through a Systematic Literature Review (SLR) using the PRISMA protocol on 25 selected articles published between 2020 and 2024. The findings indicate a shift in the dominance of evaluation tools toward the Human-Organization-Technology Fit (HOT-Fit) model and the Net Promoter Score (NPS). Key determinants of satisfaction were identified in terms of information quality, system reliability, and responsiveness of technical support
Implementation of Inter-Building Wireless Backhaul using Ubiquiti 5AC Gen2 and MikroTik
Full text linkThis study aims to implement and evaluate the performance of an inter-building wireless backhaul network using the Ubiquiti LiteBeam 5AC Gen2 integrated with a MikroTik router as a solution for internet distribution without the need to subscribe to an additional ISP service. The study is motivated by the increasing demand for high-speed and reliable network connectivity between buildings, while wired network implementations are often limited in terms of cost and installation flexibility. The research adopts an experimental approach with descriptive quantitative analysis through direct measurement of Quality of Service (QoS) parameters, including throughput, delay, and jitter. Testing was conducted under three scenarios: a direct wireless backhaul link, a WiFi network on the first floor, and a WiFi network on the second floor of the building. The results show that the wireless backhaul provides an average throughput of 87.95 Mbps for download and 48.60 Mbps for upload, with a delay of 1.19 ms. On the access network side, the achieved throughput remains sufficient for user needs, although delay and jitter increase as the number of connected devices and traffic load grows. This study concludes that the implementation of an IEEE 802.11ac-based wireless backhaul using Ubiquiti LiteBeam 5AC Gen2 and MikroTik is effective as a medium-scale inter-building connectivity solution, delivering performance that meets typical daily internet service requirements
Systematic Review of the Use of the MIT-BIH Polysomnography Database for the Detection and Classification of Sleep Disorders
Full text linkThe MIT-BIH Polysomnography Database (SLPDB) is a widely adopted benchmark for the development of automated methods for sleep disorder detection and sleep stage classification. This study presents a Systematic Literature Review of 35 articles that utilize the SLPDB, examining research focus areas, types of physiological signals employed, and the computational approaches applied. Five major methodological categories were identified: Sleep Apnea Detection, Sleep Staging, Signal Processing Enhancement, Multichannel Fusion Methods, and Interpretable Artificial Intelligence, with the first two categories being the most dominant. Four groups of physiological signals—EEG, ECG, respiratory signals, and multichannel data—form the basis for model development, where EEG is predominantly used for sleep staging and ECG for sleep apnea detection. Deep learning approaches, particularly CNNs, LSTMs, and hybrid models, are the most frequently employed techniques. Reported model accuracies range from 78% to over 99%, depending on the signal modality and modeling strategy. Future research should prioritize the development of more interpretable hybrid models and broader clinical validation to enhance reproducibility and implementation readiness
Development of Android GIS Applications for Mapping Clean Water Sources in Natural Resource Management in Disaster-Affected Areas
Full text linkThis research was motivated by the post-disaster challenges faced by the people of Sarampad Village, Cugenang District, Cianjur Regency, after the 2022 earthquake, which severely damaged vital infrastructure and disrupted access to clean water. The lack of a systematic mapping system for clean water sources highlights the urgent need for technology-based solutions to support effective and sustainable water resource management. The study employed a software development method using an Agile Programming approach, allowing iterative development and adaptation based on user feedback. Data were collected through field surveys, interviews with local communities and village officials, and direct observations of clean water source conditions. The system was designed using the Unified Modelling Language (UML), and the Android application was developed with Flutter Dart via the Visual Studio Code platform. Application functionality was tested using the black-box testing method to ensure performance reliability. The developed Android-based GIS application successfully maps and visualizes clean water sources, providing users with accurate and accessible spatial information. The system enables communities to identify the nearest clean water sources efficiently, particularly in post-disaster conditions. The findings demonstrate that integrating GIS with mobile technology can significantly improve public access to clean water information while promoting community involvement in environmental resource management. This innovation serves as a practical step toward sustainable and participatory water resource management in disaster-affected areas
Word Embedding Features to Improve Machine Learning Performance in Sentiment Analysis of the Honor of Kings Game
Full text linkThe rapid growth of social media has encouraged an increasing number of studies on sentiment analysis to better understand public perceptions and opinions. This study aims to evaluate the performance of three machine learning algorithms—Naïve Bayes, K-Nearest Neighbor (KNN), and Random Forest—in classifying user review sentiments toward the game Honor of Kings. The dataset was collected from the Google Play Store, consisting of 900 reviews. The data then underwent preprocessing steps including cleaning, case folding, tokenization, stopword removal, stemming, and sentiment labeling into positive and negative classes. Furthermore, three word embedding techniques were applied, namely Word2Vec, GloVe, and FastText, each of which was tested across the three machine learning algorithms. The experimental results indicate that the use of word embedding features significantly improves classification accuracy compared to models without embedding features. KNN combined with FastText achieved the best performance, reaching an accuracy of 87.55%, while Random Forest combined with FastText produced the lowest accuracy. FastText demonstrated superior performance due to its ability to represent words through subword information, making it more effective in handling rare vocabulary and large-scale datasets. This study confirms that combining machine learning classification methods with word embedding features plays a crucial role in improving sentiment analysis performance. Future research may focus on hyperparameter optimization, the application of more advanced preprocessing techniques, and dataset expansion to develop more robust models with better generalization capability