Jurnal Ilmu Komputer dan Informasi
Not a member yet
247 research outputs found
Sort by
Utilizing X Sentiment Analysis to Improve Stock Price Prediction Using Bidirectional Long Short-Term Memory
Full text linkThe capital market is one of the important factors that influence the national economy. However, the stock price in capital market fluctuates over time. Therefore, the investors strongly need an accurate prediction of stock price for making profitable decision. However, with the pervasive influence of the internet, investors and investment institutions have started incorporating online opinions and news, including those found on social media platforms like X. This research aims to enhance stock price prediction by utilizing X sentiment analysis. The sentiment of tweets from X related to IHSG stock price is predicted by using BERT (Bidirectional Encoder Representations from Transformers), then its result isintegrated with the historical stock price data for predicting future stock price by using BiLSTM (Bidirectional Long Short-Term Memory). The experiment results show that the RMSE and MAPE of the proposed model with sentiment analysis is decreased by 0.042 and 0.595, resepectively, compared to the model without sentiment analysis. Therefore, it can be concluded that the inclusion of X sentiment analysis in conjunction with BiLSTM succeeded in improving the performance of stock price prediction. The study's outcome is expected to be valuable for investors to make profitable decisions, leveraging the information available on social media
Indonesian License Plate Detection and Recognition System using Gaussian YOLOv7
Full text linkIn recent years, Automatic License Plate Recognition (ALPR) systems have garnered attention in computer vision research. However, practical applications face challenges such as inconsistent lighting, diverse license plate designs, and environmental variations, which increase the complexity of the task and lead to more false detections. To address these issues, we proposed Gaussian YOLOv7 for license plate detection and character recognition within ALPR systems, along with the Spatial Transformer Network (STN) for rectifying license plate orientation, aiming to enhance performance and adaptability to real-world scenarios. Additionally, we introduced a novel dataset for Indonesian ALPR systems to ensure robust detection and a balanced class distribution. Evaluation results indicate that Gaussian YOLOv7 improves precision and reduces false positives by 37.5% in the detection stage, albeit with poorer performance in other metrics. Conversely, the implementation of STN results in decreased character recognition accuracy, underscoring its limited effectiveness. Despite these challenges, Gaussian YOLOv7 excels in license plate rectification, achieving a recall of 83.8% and reducing false positives by 50.13% compared to YOLOv7. Moreover, post-processing techniques introduced by our approach further enhance precision by 5.3% and recall by 1%. Overall, our approach offers promising advancements in Indonesian ALPR systems, addressing fundamental challenges and enhancing performance
Analysis of Coding Stress Impact on Students Programming Skills with Random Forest and C4.5 Algorithms
Full text linkStudents' stress often impedes their advancement in programming, which demands logical reasoning, an understanding of algorithms, and a firm grasp of basic concepts. This research intends to pinpoint the elements that affect students' programming abilities, explore their connection to stress levels, and assess the effectiveness of the Random Forest and C4.5 algorithms in classifying data. Information was gathered through an online questionnaire involving 744 students in 2024 at various leading universities in Islamabad, Pakistan. The dataset used in this study was sourced from Kaggle, which provides insights into factors affecting students' programming performance and stress levels. The analysis utilized a Confusion Matrix and evaluation metrics like accuracy, precision, recall, and F1-Score. The analysis results indicate that the C4.5 algorithm has a higher accuracy of 68.04% compared to Random Forest, which achieved 65.54%. Additionally, C4.5 outperforms Random Forest in terms of precision, scoring 71.7% versus 65.2%. However, in terms of recall, Random Forest performs better with a score of 66.3%, while C4.5 only reaches 59.6%. This study confirms that interest in programming, debugging skills, mathematical and analytical abilities, and perceptions of programming significantly impact students' performance and stress levels. Students with strong logical abilities and adequate support demonstrate better performance and lower stress levels, whereas those with weak technical skills and negative perceptions are more vulnerable to stress, which adversely affects their performance. These findings emphasize the importance of creating a positive learning environment through interactive methods, structured problem-solving, and additional support
YOLOv11 Model as a Smart Solution for Waste Identification and Classification in Automated Waste Management System
Full text linkUrbanization and population growth present significant challenges for efficient and sustainable waste management. This research develops an IoT-based intelligent system for waste classification and management utilizing RFID technology, ESP32, a camera, an ultrasonic sensor, and the YOLOv11 object detection model. The system accurately identifies three categories of waste: organic, inorganic, and hazardous. The classification process is automated, incorporating user identification via RFID, servo-controlled bin lid operation, and capacity monitoring through an ultrasonic sensor. Data management is facilitated through a mobile application and a website, which provide user guidance and support for administrators. Test results indicate that the system achieves an average accuracy of 87.5% in the mAP50-95 evaluation, with specific accuracies of 89.0% for inorganic waste, 86.0% for hazardous waste, and 87.0% for organic waste. Despite these results, challenges remain, including object detection errors related to background interference. Future research should focus on enhancing the dataset and implementing data encryption to improve model accuracy and information security. This system demonstrates significant potential for enhancing waste management efficiency and promoting sustainable environmental practices
Attention-based Residual Long Short-Term Memory for Earthquake Return Period Prediction in the Sulawesi Region
Full text linkIndonesia, particularly the Sulawesi region, experiences significant seismic activity due to its position at the convergence of three major tectonic plates. This study seeks to construct a model for predicting earthquake return periods in the Sulawesi area by employing the Residual Long Short-Term Memory (Residual LSTM) architecture integrated with an attention mechanism. The dataset utilized originates from the United States Geological Survey (USGS), focusing on the Sulawesi Island region within the coordinates of latitude -6.184° to 2.021° and longitude 118.433° to 125.552°, spanning the years 1975 to 2024. The research methodology is structured into three primary phases: (1) data collection and preprocessing, including data cleaning, missing value handling, and normalization, (2) exploratory data analysis to understand seismic data characteristics, and (3) development of the Residual LSTM model with an attention mechanism. The evaluation results show excellent model performance with Train Loss 0.0090, Test Loss 0.0091, Training MAE 0.0698, Testing MAE 0.0717, Training RMSE 0.0947, Testing RMSE 0.0951, and stable Huber Loss of 0.0045 for both training and testing data. The implementation of residual connections successfully addressed the vanishing gradient problem, while the attention mechanism enhanced prediction interpretability. The small discrepancy between the training and testing metrics confirms the model's robust generalization ability, indicating its strong potential for applications in predicting earthquake return periods
Transformative Insights into Corrosion Inhibition: A Machine Learning Journey from Prediction to Web-Based Application
Full text linkThis study focuses on the exploration and evaluation of machine learning (ML) models to analyze expired pharmaceutical data for their potential use as corrosion inhibitors. Additionally, the entire modeling process is integrated into a user-friendly platform through a Streamlit service-assisted corrosion inhibitor website, facilitating broader accessibility and practical application. The models are trained offline to ensure accurate performance, eliminating the need for users to retrain the models themselves. This approach simplifies the user experience by offering a ready-to-use prediction service directly on the website platform. Among the various ML models implemented, XGB demonstrated the highest performance with an R2-score of 0.99999999. Given that many chemists are not familiar with informatics coding, the researchers developed a Streamlit-based website that includes tools to customize the models. The end product of this work is a corrosion inhibitor experimentation tool that eliminates the need for users to code, making advanced ML techniques accessible to a broader audience within the chemistry community
The Conceptual Design e-Wallet for Rupiah Digital
Full text linkThis research study the advancement of Central Bank Digital Currencies (CBDCs) spurred by fi-nancial technology progress. It focuses on Rupiah Digital, Indonesia's CBDC initiative led by the Bank of Indonesia (BI). The study explores the technical aspects of Wholesale and Retail Digital Rupiah, proposes an e-wallet system for seamless digital transactions in related to blockchain technology, specifically Permissioned Distributed Ledger Technology (DLT). The objective of this research to provide recommendations to BI regarding appropriate e-wallet conceptual design based on study literature review (LR) methods and qualitative research method by conducting interviews throughs forum group discussion (FGD) and e-mail with leading economic (banks), legal (BI and Government), and technical experts (banks, academic expert on this field, BI and Government) to get reviews and input regarding the e-wallet conceptual design that was proposed. As result, we recommended the architecture for Rupiah Digital using Hyperledger Fabric blockchain with two-tiered distribution and user layer backed by digital token using ID on mobile apps to enhance the security of the system. The FGD with experts and executor result in approval on those conceptual design to be part of the option on development of CBDC in Indonesia
Comparing ASM and Learning-Based Methods for Satellite Image Dehazing
Full text linkRecent advancements in optical satellite technologies have significantly improved image resolution, providing more detailed information about Earth's surface. However, atmospheric interference, such as haze, is still a major factor in image capture. The interference results in visibility degradation of the acquired images, hindering computer vision tasks. Numerous studies have proposed various methods to recover haze-affected regions in satellite images, highlighting the need for more effective solutions. Motivated by this, this paper compares different atmospheric dehazing methods, including Atmospheric Scattering Model (ASM)-based and deep learning-based. The results show that SRD is the best ASM-based method, with a PSNR value of 19.09 dB and an SSIM of 0.908. Among deep learning models, DW-GAN achieves the best restoration results with a PSNR value of 26.22 dB and an SSIM of 0.959. SRD offers faster inference times, but still suffers from residual haze and noticeable color degradation compared to DW-GAN. In contrast, DW-GAN provides a more complete haze removal at the cost of higher computational demands than ASM-based methods
A Hybrid Vision Transformer Model for Efficient Waste Classification
Full text linkThe rapid and accurate sorting of municipal waste is essential for efficient recycling and sustainable resource recovery. Most existing AI solutions focus only on four common materials (plastic, paper, metal, and glass), overlooking many other routinely encountered waste types and losing accuracy when applied to the mixed waste compositions seen in operational environments. We introduce HR-ViT, a hybrid network that combines ResNet50 residual blocks, which capture fine-grained local cues, with Vision Transformer global self-attention. Trained on a balanced six-class benchmark of about 775 images per class (plastic, paper, organic, metal, glass, batteries), HR-ViT attains 98.27 % accuracy and a macro-averaged F1-score of 0.98, outperforming a pure ViT, VT-MLH-CNN, and Garbage FusionNet by up to five percentage points in both metrics. Gains arise from selective fine-tuning of the last ten ResNet layers, lightweight ViT hyper-parameter optimisation, and targeted data augmentation that mitigates cluttered backgrounds, uneven lighting, and object deformation. These results show that hybrid attention-residual architectures provide reliable predictions under complex imaging conditions. Future work will extend the method to multi-object scenes and domain-adaptive deployment in smart-city recycling systems
Comparative Evaluation of Database Systems for High-Volume Seismic Prediction Data Management in Real-Time Applications
Full text linkThe Earthquake Early Warning System (EEWS) plays a pivotal role in mitigating structural damage and minimizing casualties by issuing alerts prior to the arrival of destructive seismic waves (S-waves), through the detection of the earlier and faster P-waves. The operational effectiveness of EEWS depends not only on the accuracy of its predictive algorithms but also on the efficiency of the underlying data storage and management infrastructure. This study presents a comparative evaluation of three data storage approaches i.e. MongoDB, MongoDB with sharding, and InfluxDB, as well as the MiniSEED (mseed) binary format, with a focus on their performance in managing real-time seismic prediction data. Benchmarking was conducted based on two key metrics: Input/Output Operations Per Second (IOPS) and data throughput. The results indicate that both MongoDB and InfluxDB offer strong performance in high-ingestion scenarios, with MongoDB demonstrating higher IOPS, while InfluxDB exhibits better scalability and consistency as data volume increases. Conversely, the mseed format achieves exceptionally high throughput due to its flat-file structure but lacks the responsiveness and query capabilities required for real-time analytics. These findings suggest that MongoDB and InfluxDB are well-suited for integration into scalable EEWS infrastructures, offering a balance between performance and flexibility. Future work will extend this evaluation to larger-scale datasets and alternative architectures such as data lake systems to improve disaster response readiness