JUTI: Jurnal Ilmiah Teknologi Informasi
Not a member yet
407 research outputs found
Sort by
A Dual-Network iTransformer Model for Robust and Efficient Time Series Forecasting
Full text linkTime-series forecasting plays a crucial role in various fields, including economics, healthcare, and meteorology, where accurate predictions are essential for informed decision-making. As data volume and complexity continue to grow, the need for efficient and reliable forecasting methods has become more critical. iTransformer, a recent innovation, improves interpretability while effectively handling multivariate data. In this study, the author proposes Dual-Net iTransformer, a novel approach that integrates iTransformer with a dual-network framework to enhance both accuracy and efficiency in time-series forecasting. This research aims to evaluate and compare the performance of traditional methods, iTransformer, and Dual-Net iTransformer, highlighting the advantages of the proposed model in improving forecasting outcomes
Exploring The Role of Augmented Reality in Education: Systematic Literature Review
Full text link The development of digital technology drives innovation in education, one of which is through the implementation of augmented Reality (AR), which increases interactivity and understanding of abstract concepts in learning. This study employs a Systematic Literature Review (SLR) with the PRISMA method to analyze the implementation of AR in education. Of the 3,225,372 articles reviewed, 30 journals met the research criteria, with Marker-Based Tracking as the most commonly used AR method because of its stability and accuracy. The study results showed that AR increases students\u27 interactivity, facilitates understanding of abstract concepts, increases student engagement, improves information retention and memory, facilitates simulation and practice, develops creativity and collaboration, adapts learning to individual needs, and improves cost and resource efficiency, although it still faces challenges in the infrastructure and technical skills of teachers. Therefore, further development in AR applications at various education levels is recommended to improve understanding and adaptation to scientific developments.
Keywords: Augmented Reality, Learning Media, Marker-Based Tracking
Evaluating Object Collection In Emergency Simulations Using Virtual And Augmented Reality
Full text linkVirtual Reality (VR) and Augmented Reality (AR) are two technologies that have received significant attention in recent years. While both hold immense potential, they offer distinct ways for users to interact with digital content and their physical surroundings. This research aims to evaluate the interaction between users and a collection of objects in both VR and AR settings. To achieve this, a user study was conducted with 24 participant using Meta Quest 3 headset to run simulation in both environments. The study focused on tasks related to object collection and emergency management while utilizing combination of objective and subjective metrics to evaluate user interactions in both VR and AR environments. Despite the relatively close scores for both result, research shows that participants prefer AR for emergency simulations over VR. Even considering participants\u27 first-time use of the applications, AR remains more popular, supported by lower symptom rates reported in the sickness than VR. Additionally, participants tended to focus more on collecting small objects, though VR users often forgot these items, while medium-sized objects were more frequently overlooked in AR. Although VR users experienced more human errors related to collisions with real objects, the overall impact on immersion during simulations was not significant enough to favor one technology over the other. Based on this result, it can be said that while VR is better for showing immersion, it is generally better for a first-time user to engage in AR first since it will give less incidence of virtual sickness
Gated Recurrent Unit Based Predictive Modeling for Dynamic Obstacle Avoidance in Autonomous Aerial Vehicles
Full text linkThe entry of Autonomous Aerial Vehicles (AAVs) has reshaped multiple industries through novel solutions such as transport, monitoring, and deliveries. Nevertheless, the existence of dynamic operating environments, and the unpredictability of barrier emergence, constitutes a complicated path planning challenge that is difficult to cope with. Current methods of dynamic obstacle avoidance, e.g. Recurrent Neural Networks (RNNs) and Long Short-term Memory (LSTM) networks, accomplished the task and became the essential part of AAV navigation systems development. These techniques may work, but they have a disadvantage of being slow in processing and less energy efficient, which are important for a real-time operation and for a mission which lasts for a long time. The purpose of the research is to fill up the identified gaps by introducing a GRU-based predictive model for dynamic obstacle avoidance in AAV’s. While the previous models concentrate on the improvement of reaction time and energy consumption without the degradation of computational efficiency, the recent GRU model is particularly designed for such purpose. It is realized through a streamlined design that facilitates rapid and precise object trajectory predictions, thus, making AAVs be able to rethink their paths in advance of any obstacles lurking. We show that the RNN-based GRU model is benchmarked significantly better than the RNN and LSTM models in simulated settings. In the Eco mode, the model GRU responded in 0.35 seconds in low-speed and its energy consumption never exceeded 130 units even in the high-speed scenarios with maximum load. Path efficiency was preserved and the path length was kept to the minimum in most cases, which indicates the model\u27s capability in finding the most direct paths. Additionally, computer loads were at a tolerable level, thus further showing the applicability of this model for systems on-board having inducted limits for their processing capabilities. GRU- based model comes out as a robust and economical technique for the obstacle avoidance, giving a potential solution to the critical problems of AAVs
Gambling Comments Detection on Youtube: A Comparative Study of Tree-Based Boosting, LSTM and GRU Models
Full text linkThe exponential growth of online gambling in Indonesia poses significant socio-economic challenges, particularly affecting vulnerable populations through sophisticated digital marketing strategies targeting social media platforms. This study addresses the critical need for automated detection systems to identify gambling-related content in YouTube comments. We scraped and manually labeled 11,673 comments from diverse YouTube videos, creating an extremely imbalanced dataset with gambling comments representing only 10% of the total data. Multiple machine learning approaches were developed and evaluated, comparing traditional gradient boosting methods (LightGBM, XGBoost, CatBoost) using TF-IDF features against deep learning models (LSTM & GRU) with Word2Vec embeddings. The experimental results demonstrate that gradient boosting methods significantly outperform deep learning approaches in generalization capability. LightGBM achieved the highest holdout F1-score with balanced precision (0.8912) and recall (0.8886), while XGBoost followed closely with comparable performance. In contrast, deep learning models exhibited severe overfitting, with GRU and LSTM showing excellent test performance but drastically reduced holdout recall (0.5022 and 0.4844, respectively). The findings indicate that the dataset size was insufficient for deep learning approaches to learn generalizable representations effectively. For practical deployment in YouTube gambling content detection, gradient boosting methods are recommended due to their superior performance with limited, imbalanced datasets
Network Intrusion Detection System with Time-Based Sequential Cluster Models using LSTM and GRU
Full text linkTechnological development and the growth of the internet today have a positive and revolutionary impact in various areas of human life, such as banking, health, science, and more. The presence of Open Data and Open API also facilitates the exchange of data and information between entities without the restrictions imposed by different regions and geographical areas. However, information openness not only has a positive impact but also makes data vulnerable to data theft, viruses, and various other types of cyber attacks. The large-scale data exchange that occurs across the network poses a challenge in detecting unusual activity and new cyber attacks. Therefore, the existence of an Intrusion Detection System (IDS) is urgently essential. The IDS helps system administrators detect cyber attacks and network anomalies, thus minimizing the risk of data leaks and intrusions. The research developed a new approach using time-based sequential clustered data sets in the Long Short Term Memory (LSTM) and Gated Recurrent Unit (GRU) models. This IDS model was implemented using the CIC-IDS 2018 data set, which has more than 4 million data lines. The capabilities and uniqueness of the LSTM and GRU models are used to classify and determine various attacks in IDS based on sequential data sets ordered by time and clustered according to the destination ports and protocols, such as TCP and UDP. The model was evaluated using the accuracy, precision, recall, and F-1 scores matrix, and the results showed that the time-based sequential clustered models in LSTM and GRU have an accurities of up to 97.21%. This suggests that this new approach is good enough to be applied to the future IDS models
Garch Model Hybridization With Feed Forward Neural Network Algorithm Approach For Predicting The Volatility Of The Composite Stock Price Index
Full text linkStock market volatility is a crucial indicator in measuring investment risk and influencing investor decision-making, where proper understanding of volatility movements can help investors optimize their investment portfolios. Time series data from the stock exchange shows complex heteroscedasticity characteristics, where volatility levels can change dynamically over time, creating distinct challenges in modeling and prediction. The implementation of the hybrid model is carried out by integrating the advantages of both models, where GARCH is used to capture volatility clustering characteristics, while FFNN is utilized to capture complex non-linear patterns in the data.
By using evaluation of several comprehensive error measurement metrics, including Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE), to ensure model reliability in various aspects of prediction. The use of the GARCH-FFNN hybrid model is expected to provide more accurate volatility predictions compared to using GARCH or FFNN models separately, with potential improvements in prediction accuracy and adaptability to changing market conditions. These findings provide important contributions to stock market volatility modeling and can serve as a reference for investors, portfolio managers, and financial practitioners in making better investment decisions, as well as paving the way for the development of more sophisticated volatility prediction models in the futur
Deep Metric Learning with Different Distance Metrics for Enhanced Classification Model in Typing Style
Full text linkWriting can be a powerful and unique medium of self-expression for every individual. Therefore, we propound a deep metric learning technique to acquire the vector representation of text, aiming to enhance the performance of deep learning classification models in typing style classification. The study also compared the effect of text pre-processing and distance metrics on model performance using tweet data from six different Twitter users. The outcomes of the study showed that the model without text pre-processing and with deep metric learning using the Cosine distance metric had the optimal result with an accuracy of 0.79, compared to the deep learning model with a categorical cross-entropy loss function which only had an accuracy of 0.76. Additionally, the model with text pre-processing also produced a good performance, with an accuracy of 0.63 using the deep metric learning approach and Cosine distance metric, and an accuracy of 0.64 using deep learning classification with a categorical cross-entropy loss function
Audio Feature Analysis and Selection for Deception Detection in Court Proceedings
Full text linkDeception detection is a method to determine whether a person is lying or not. One lie detector is a polygraph that measures human physiology, such as pulse and blood pressure. However, polygraphs have a problem in that they cannot be measured based on human psychology, such as speech and intonation. Therefore, audio deception detection is required, and this can be measured based on human psychology. This research will extract audio features, such as the Mel Frequency Cepstral Coeffi-cient (MFCC), Jitter, Fundamental Frequency (F0), and Perceptual Linear Prediction (PLP), from the Real-Life Trial dataset, which comprises 121 audio data. From the extraction results in the form of numerical data totaling 6387 features, various feature-selection methods are employed, such as Feature Importance (FI), Principal Component Analysis (PCA), Information Gain, Chi-Square, and Recursive Feature Elimination (RFE). After feature selection, the selected features are input to machine learning models, such as random forest and support vector machine (SVM). After model testing, metrics such as accuracy, precision, recall, and F1 score were evaluated, as well as statistical evaluation, to assess the developed model. Results from this experiment show that the deception detection model is improved after a feature selection process to reduce irrelevant features. Comparing the accuracy, Chi-Square achieves a significantly higher result, reaching up to 92% with an improvement of 24.32%, surpassing the SVM model\u27s accuracy of 67.57% before feature selection. In contrast, the RFE technique yielded the best accuracy of 86%, with an increase of 13.52%, building upon its baseline accuracy of 72.97%
DDoS Mitigation in Kubernetes: A Review of ExtendedBerkeley Packet Filtering and eXpress Data Path Technologies
Full text linkKubernetes, as a widely adopted container orchestration platform, is increasingly targeted by sophisticated cyber threats, including Distributed Denial of Service (DDoS) attacks, which can severely compromise the stability, availability, and operational integrity of Kubernetes clusters by overwhelming the cluster’s control plane, disrupting pod scheduling, or saturating network resources. Emerging Linux kernel technologies, such as the Extended Berkeley Packet Filter (eBPF) and eXpress Data Path (XDP), offer innovative and efficient solutions to mitigate these challenges by enabling high-performance packet filtering, real-time traffic monitoring, and advanced intrusion detection directly within the kernel. These capabilities help reduce latency, enhance resource efficiency, and strengthen the security posture of modern cloud-native environments. This review explores advancements in network security by examining the integration of eBPF and XDP for defending Kubernetes environments against DDoS attacks. By analyzing existing studies and identifying their limitations, this review highlights the potential of these technologies to establish efficient, scalable, and adaptive mitigation frameworks. The insights gained from this research can guide the development of robust security policies tailored for modern containerized infrastructures