JOIV : International Journal on Informatics Visualization
Not a member yet
786 research outputs found
Sort by
Exploring the Capabilities of GPT Models in Drafting Course Assessments Based on Bloom’s Taxonomy
Get PDFThe application of Generative Pre-trained Transformer (GPT) models is significantly essential in automating drafting course assessment based on Bloom’s Taxonomy, specifically GPT-3.5-turbo, GPT-4, and GPT-4o. Therefore, this study aimed to explore the interaction between Artificial Intelligence (AI) models and educational content using refined prompt engineering methods to enhance the accuracy and relevance of the generated questions. For the investigation, the processing 146 Course Learning Outcomes (CLOs) method was applied through each model using OpenAI Application Programming Interface (API). Metrics such as 'Accuracy', 'Precision', 'Recall', and 'F1 Score' were used to assess the performance of each model. The results showed that GPT-4 was suitable for complex course assessments, showing superior performance in delivering detailed and precise responses. A cost-effective solution was obtained using GPT-3.5-turbo for generating simpler course assessment, while GPT-4o provided a middle ground, balancing cost, and performance. The results showed the potential of AI to reduce the administrative burden on instructors by streamlining the creation and refinement of course assessments. The enhancement of course assessments was also facilitated by automation, thereby supporting more adaptive questions. The potential for broader AI integration into educational practices promised a transformative impact on traditional course assessment drafting methods, enabling more dynamic and educational experiences. Moreover, further studies were recommended to explore the ethical dimensions of AI in education, the ability to handle diverse tasks, as well as assess the long-term impacts on learning outcomes and educational equity
Combination of Multidistance Signal Level Difference and Time Domain Features for Epileptic Seizure Classification
Get PDFEpileptic seizures are neurological disorders characterized by abnormal electrical activity in the brain, causing a series of seizures or episodes of temporary loss of consciousness. This research aims to develop a method of detecting and classifying epileptic seizures using one-dimensional EEG signals with the Multidistance Signal Level Difference (MSLD) approach and time domain feature extraction. The goal is to improve accuracy in distinguishing normal, interictal, and ictal conditions in EEG signals. The dataset from Bonn University consists of one-dimensional EEG signals that include normal, interictal, and ictal states. The analysis method includes extracting time domain features from EEG signals, such as Integrated EMG (IEMG), Mean Absolute Value (MAV), and others. The next step is the application of three classification algorithms, namely linear SVM, quadratic SVM, and cubic SVM, to classify the three conditions. Testing is done by measuring the accuracy of the classification results. The results of this study show that by using 14-time domain features and the MSLD approach, the most significant classification accuracy achieved was 98.7%. This result demonstrates the effectiveness of the proposed method in distinguishing normal, interictal, and ictal conditions. This research provides a foundation for further study in developing EEG signal classification analysis models. Future research can expand the scope by considering larger datasets, using more sophisticated feature extraction techniques, and exploring more complex classification algorithms to improve the accuracy and reliability of the model in real-world applications, particularly in the medical field for the diagnosis of epileptic seizures
Attendance System Leveraging Haar Cascade Detection And CNN-Based Facenet Recognition Technology
Get PDFThe objective of this research is to investigate face identification methods in the context of employee recognition as a solution to the problem of attendance that still uses manual methods or applications without identity validation. The main goal is to achieve optimal accuracy and consistency in the identification process using Convolutional Neural Networks (CNN) with FaceNet and Haar Cascade. This research focuses on the challenge of managing employee attendance, particularly for those who are working remotely, which can be vulnerable to fraudulent activity. The proposed solution combines facial recognition to enhance identity verification, attendance tracking, and assist companies in achieving their goals. The study employed a dataset of 1,050 employee face data and divided it into three scenarios for training and testing ratios: the first scenario (80:20), the second scenario (70:30), and the third scenario (60:40). The results indicate that the model in the first scenario had the highest accuracy value of 98% and outperformed the models in the second and third scenarios in terms of precision, recall, and f1-score, with values of 98.60%, 98.70%, and 98.60%, respectively. The results indicate that the model used in the first scenario is the most effective in classifying predicted cases and consistently predicting employee identification. Based on these findings, we recommend implementing suggestions such as adding datasets and analyzing important classes to improve the accuracy and generalization of face identification models in the context of employee recognition. Combining facial recognition improves identity verification and attendance tracking, making it easier for companies to manage employee attendance with greater effectiveness.
Classification of Coral Images Using Support Vector Machine with Gray Level Co-Occurrence Matrix Feature Extraction
Get PDFThis research developed a coral image classification method using Support Vector Machine (SVM) with Gray Level Co-occurrence Matrix (GLCM) feature extraction to improve the accuracy of coral reef condition monitoring. Coral images were collected in the waters of Sangihe Islands Regency and labelled by experts for healthy, unhealthy, and dead categories. Preprocessing included cropping, background removal, sharpening, and image normalization. GLCM feature extraction was performed with a distance of 1, 2, and 3 pixels and directions of 0°, 45°, 90°, and 135°. SVM uses Linear, Radial Basis Function, and Polynomial kernels with parameters set in a grid. The results indicate that the polynomial kernel with parameters C=10, degree=3, and gamma=1 achieves the highest accuracy, at 91.85%. Oversampling increased the accuracy by 2.17%, while feature selection by boxplot and model-based decreased the accuracy by 0.8% and 0.2%, respectively. On the other hand, feature selection using correlation analysis significantly decreased accuracy by 16.11%. These findings significantly contribute to coral reef conservation by offering a more accurate and efficient classification method. This method enables better and timely monitoring of coral reef conditions, thus supporting more effective conservation interventions. Integrating these research results into IoT systems can improve overall coral reef monitoring and conservation efforts
Development of Augmented Reality Media for Local Wisdom Learning
Get PDFIndonesia, as an archipelago, has much local wisdom. Teaching local wisdom materials in Indonesia so far tends to use images or videos that are not interactive. This medium has limitations in terms of student involvement. Therefore, it is necessary to develop learning media that can effectively display local wisdom from various cultures. This research uses the Research and Development method to develop Augmented Reality applications. Augmented Reality Local Wisdom: This research developed local wisdom in Indonesia, including the Batak, Javanese, Sundanese, and Betawi tribes. Material culture encompasses traditional clothing, vernacular architecture, indigenous weapons, and regional motifs. The research population comes from the provinces of North Sumatra, Jakarta, West Java, and Yogyakarta. This study's research product involves developing a project-based learning model grounded in local wisdom and supported by Augmented Reality. The development results show that using Augmented Reality in learning, particularly through the project-based learning model, is beneficial because it helps students describe material culture more effectively. This study suggests that students can become more engaged and interactive in their learning. Additionally, using Augmented Reality, students can explore culture more contextually, enhance their problem-solving skills, and reduce the time and costs associated with learning about local culture. The use of AR in local wisdom-based learning could be an innovative solution to enhance the quality of education and cultural preservation in the digital age
Secondary Structure Protein Prediction-based First Level Features Extraction Using U-Net and Sparse Auto-encoder
Get PDFProtein secondary structure prediction (PSSP) is an important challenge in bioinformatics. Existing methods for PSSP are generally divided into three categories: neighbor-based, model-based, and meta-estimator-based methods, each using supervised or supervised learning methods model-based are often neural networks, hidden Markov models are available; they support vector machines and other machine learning techniques based on multiple sequence alignments and evolutionary data from increasingly large protein databases. This paper presents a powerful machine learning approach for PSSP, which is a new feature extraction method using sparse autoencoders to identify new protein features. The sparse autoencoder efficiently identifies new features in the training data and provides an accurate prediction of occurrences. Two machine learning methods are used: unsupervised learning methods based on sparse auto-encoders and semi-supervised learning methods using deep learning methods. Experimental results show that the deep learning method gets 86.719% accuracy on the test set, while the unsupervised pretraining method gets 85.853% accuracy on the training set after being improved by surface propagation. Fine-tuning and layer-wise pretraining significantly improve the performance of the proposed model. The results show that the deep learning method achieves an accuracy of 86.7% in the training set and 71.4% in the test set. In comparison, Sparse Autoencoders alone achieved an accuracy of 67%, demonstrating the effectiveness of the combination of these methods. This study highlights the role of advanced deep learning techniques in PSSP accuracy. Future research should consider using big data, exploring deep learning algorithms, and refining optimization methods to further encourage predictive performance in bioinformatics
Elderly Acceptance of Autonomous Vehicles in Malaysia: An Extended Technology Acceptance Model with Multidimensional Trust and Perceived Risk
Get PDFThe emergence of autonomous vehicle technology is propagated to address the needs of the elderly and reduce other negative externalities brought by transportation mobility. However, these benefits would not be realized without widespread acceptance. This research aimed to investigate the factors influencing the acceptance of autonomous vehicles among the elderly in Malaysia. Building on the technology acceptance model with multidimensional trust, perceived risks, and technology anxiety, a sample of 289 elderly people within Klang Valley are included in the model estimation. Results show that the mediating roles of perceived ease of use, perceived usefulness, and attitude between trust in institutions and acceptance are not supported. On the other hand, performance trust indirectly affects acceptance through perceived ease of use, usefulness, and attitude. The multidimensional perceived risks, including perceived performance risk, privacy risk, and technology anxiety, did not support the direct effect on acceptance of autonomous vehicles. These findings validate the role of multi-dimensional trusts and perceived risks in accepting autonomous cars. Trust and perceived risk in autonomous vehicles evolve; thus, a longitudinal study is recommended for future studies to understand better the elderly's acceptance of autonomous vehicles in Malaysia as the industry matures. The findings also provide important insight into industry players who design transport policies. Building trust in autonomous cars focusing on reliability and trustworthiness is vital for widespread acceptance, particularly among the elderly
Visualizing and Mapping Research on Consumers’ Credit and Payment: A Bibliometric Analysis
Get PDFThe COVID pandemic that hit the world has had an extraordinary effect, especially on business survival. Many businesses have experienced problems with declining market share and revenues, and have even gone bankrupt due to the declining purchasing power of people worldwide. Demand and supply face various issues that exacerbate economic conditions. From the consumer side, the income decline led to an increase in bad loans, as it altered consumption patterns. Meanwhile, from the supply side, the company suffered revenue losses due to bad loans, which negatively impacted its performance. This research aimed to analyze preliminary research on credit and consumer payment and identify areas for future investigation. Data were collected from 57 journals published between 2019 and 2023 using the ProQuest search engine. Furthermore, the data obtained were analyzed to eliminate journals without quartiles that have been discontinued using the ScimagoJR website. Mendeley software was employed to identify pertinent keywords. At the same time, Vos Viewer was used to determine various variables and research methodologies, resulting in five clusters based on occurrence and the strength of links. The results indicated that most research in China used the technology acceptance model and regression modeling analysis. Based on the analysis results using Vos Viewer, research gaps remained evident, indicating a need to explore a range of variables, methods, and analytical tools. According to the clusterization in the discussion section, this research offers implications for researchers interested in consumer behavior to capitalize on opportunities
Development of Conventional Lathe Machine Manual User by Using Augmented Reality Frameworks
No full textMachining is one of the familiar subjects in the field of Technical and Vocational Education and Training (TVET) and has been offered at several Vocational Colleges and Institutes of Higher Education (IPT) throughout Malaysia. However, the level of dominance is limited to a handful of students in understanding the learning content and achieving learning outcomes at the end of the course's teaching and learning process. Therefore, this research intends to design and develop a machine manual using an interactive multimedia concept characterized by Augmented Reality (AR). The method of creating forms and developing interactive multimedia routinely uses the Analysis, Design, Development, Implementation, and Evaluation (ADDIE) model as a reference model and guideline for implementing learning. The research instruments used were product development expert review forms and student investigation questionnaires. The research respondents consisted of 80 TVET students from Universiti Tun Hussein Onn Malaysia (UTHM) and Tanjung Piai Vocational School. The data obtained is collected and analyzed periodically using statistical-based software. An evaluation is conducted on the product's design, form, content, and functionality. The results of the analysis on the use of interactive multimedia concepts indicate that the average minimum standard for all variables exceeds 3.25, which is interpreted as Highly Acceptable for the Use of Multimedia-Based Learning. Three experts in the field of multimedia and engineering agree that the product developed has a shape that matches the design and can function effectively. Overall, the research found that the design form, content, and functionality of conventional interactive machines can enhance students' visualization abilities in the teaching and learning process, as well as improve their skills when practicing with the devices
Predictive Analytics for Employability in Malaysian TVET with a Hybrid of Regression and Clustering Methods
Get PDFGraduate employability remains a high concern for Technical and Vocational Education and Training (TVET) institutions, particularly within Malaysia’s Technical University Network (MTUN), where producing industry-ready graduates is a central goal. While machine learning has transformed fields like healthcare and finance, its application in vocational education remains underexplored—particularly for employability prediction. This study addresses this gap by hybridizing decision trees and clustering to uncover non-linear patterns in student survey data. Guided by Human Capital Theory and SERVQUAL, which inform variable selection (e.g., technical skills as productivity investments), this study integrates multiple linear regression, decision tree regression, and K-Means clustering to identify significant predictors and uncover latent student groupings. Using a publicly available dataset of Likert-scale responses from MTUN students, technical skills and supervisory support consistently emerged as the most impactful employability predictors. Communication showed moderate influence, while training delivery and problem-solving exhibited variable effects depending on the modelling approach. Unlike regression, decision trees revealed non-linear interaction thresholds. For example, students with SVR < 3.5 and TS < 4.0 had 40% lower employability scores, suggesting targeted mentoring could yield disproportionate improvements. Clustering revealed three distinct student profiles, which could support data-driven interventions. This hybrid framework demonstrates the potential for integrating machine learning into institutional analytics for proactive support of employability