Computer Science and Information Technologies (E-Journal)
Not a member yet
149 research outputs found
Sort by
Implementation of IoT-based water quality monitoring instruments in cantang grouper cultivation ponds
Full text linkGrouper fish farming in Indonesia has great potential, but water quality management remains a challenge. Manual monitoring at hatchery D-Marine aquaculture struggles to detect sudden changes, risking mass mortality. This study developed an IoT-based water quality monitoring system using an ESP32 microcontroller, DS18B20 temperature sensors, pH sensors, dissolved oxygen (DO) sensors, a micro-SD card, an organic light emitting diode (OLED) display, and the Ubidots platform. The methodology involved device design, sensor calibration, and field testing. Calibration showed sensor accuracy above 90%. Field tests recorded water temperatures of 26.84 °C (tank 1) and 27.74 °C (tank 2), with pH values of 6.73 and 6.87, which did not meet Indonesian national standard (SNI) standards. Data transmission to Ubidots had a 95% packet delivery ratio (PDR) for device 1 and 97% for device 2. The system successfully provided real-time water quality data, supporting effective farm management. However, improvements to the dissolved oxygen sensor and an automatic control system are needed for better stability and efficiency
Javanese and Sundanese speech recognition using Whisper
Full text linkAutomatic speech recognition (ASR) technology is essential for advancing human-computer interaction, particularly in a linguistically diverse country like Indonesia, where approximately 700 native languages are spoken, including widely used languages like Javanese and Sundanese. This study leverages the pre-trained Whisper Small model an end‑to‑end transformer pretrained on 680,000 hours of multilingual speech, fine tuning it specifically to improve ASR performance for these low resource languages. The primary goal is to increase transcription accuracy and reliability for Javanese and Sundanese, which have historically had limited ASR resources. Approximately 100 hours of speech from OpenSLR were selected, covering both reading and conversational prompts, the data exhibited dialectal variation, ambient noise, and incomplete demographic metadata, necessitating normalization and fixed‑length padding. with model evaluation based on the word error rate (WER) metric. Unlike approaches that combine separate acoustic encoders with external language models, Whisper unified architecture streamlines adaptation for low‑resource settings. Evaluated on held‑out test sets, the fine‑tuned models achieved Word Error Rates of 14.97% for Javanese and 2.03% for Sundanese, substantially outperforming baseline systems. These results demonstrate Whisper effectiveness in low‑resource ASR and highlight its potential to enhance transcription accuracy, support language preservation, and broaden digital access for underrepresented speech communities
Matrix inversion using multiple-input multiple-output adaptive filtering
Full text linkA new approach for matrix inversion is introduced. The approach is based on vector representation of multiple-input multiple-output (MIMO) channel matrix, in which the channel matrix is described by a linear combination of channel vectors weighted by their respective system inputs. The MIMO system output is then fed into a bank of adaptive filters, wherein the response of a given adaptive filter is iteratively minimized to match its output to the given system input. In doing so, adaptive filters equalize the impact of respective channel vectors on the MIMO channel output, while simultaneously orthogonalizing themselves from all other channel vectors, forming the channel matrix inverse. The method demonstrates satisfactory convergence and accuracy in Monte Carlo simulations conducted with varying signal-to-noise ratios (SNRs) and matrix conditioning scenarios. The suggested approach, by virtue of its adaptable characteristics, can also be employed for time-varying linear equation systems
Power of analytic tools in Oxygen Forensic® Detective based on NIST cybersecurity framework
Full text linkThe National Institute of Standards and Technology (NIST) cybersecurity framework is a systematic approach for assessing and improving cybersecurity procedures in digital investigations. Oxygen Forensic® Detective is a digital forensic software that integrates multiple analytic tools to assist investigators in extracting valuable insights from digital evidence. The analytic tools, including timeline, social graph, image categorization, facial categorization, maps, data search, key evidence, optical character recognition, statistics, and translation, assist investigators in thoroughly analyzing digital artifacts, establishing connections, and accurately classifying images with precision and effectiveness. By incorporating these analytical resources into Oxygen Forensic® Detective, a comprehensive strategy is established to effectively combat cyber threats. The NIST cybersecurity framework is incorporated into the tool, offering a methodical approach to identifying and reducing cybersecurity risks. Law enforcement agencies can enhance the productivity and effectiveness of their forensic methodologies by implementing these advanced technologies. This can result in successful prosecutions and improved cybersecurity practices. Overall, the utilization of analytical tools in criminological inquiries has experienced a substantial rise in the contemporary digital era
Retraction Notice: Quay crane assignment in container terminals using a genetic algorithm
No full textPorts are essential for international trade, connecting production areas to consumer markets. However, port operations often face delays, disrupting supply chains and increasing transit times. To address this, mathematical modeling, particularly through genetic algorithms, offers a solution for optimizing processes like container unloading. This paper presents a model predicting and optimizing unloading times by considering factors such as crane types, schedules, and environmental conditions. Focusing on the Casablanca port, the model addresses scheduling for two gantry and two mobile cranes, treating each bay as a unique task handled by one crane type to avoid conflicts. Using genetic algorithms, the goal is to create efficient schedules that minimize waiting times and maximize crane utilization. The expected outcome is a detailed timetable enabling effective gantry crane use or simultaneous multi-crane operations, enhancing unloading efficiency. This approach can be adapted to other ports with similar challenges, highlighting the model's broader applicability
Detection of android malware with deep learning method using convolutional neural network model
Full text linkAndroid malware is an application that targets Android devices to steal crucial data, including money or confidential information from Android users. Recent years have seen a surge in research on Android malware, as its types continue to evolve, and cybersecurity requires periodic improvements. This research focuses on detecting Android malware attack patterns using deep learning and convolutional neural network (CNN) models, which classify and detect malware attack patterns on Android devices into two categories: malware and non-malware. This research contributes to understanding how effective the CNN models are by comparing the ratio of data used with several epochs. We effectively use CNN models to detect malware attack patterns. The results show that the deep learning method with the CNN model can manage unstructured data. The research results indicate that the CNN model demonstrates a minimal error rate during evaluation. The comparison of accuracy, precision, recall, F1 Score, and area under the curve (AUC) values demonstrates the recognition of malware attack patterns, reaching an average of 92% accuracy in data testing. This provides a holistic understanding of the model's performance and its practical utility in detecting Android malware
Artificial intelligence-powered robotics across domains: challenges and future trajectories
Full text linkThe rise of artificial intelligence (AI) in robotic systems raises both challenges and opportunities. This technological change necessitates rethinking workforce skills, resulting in new qualifications and potentially outdated jobs. Advancements in AI-based robots have made operations more efficient and precise, but they also raise ethical issues such as job loss and responsibility for robot decisions. This study explores AI-powered robotics in both of their challenges and future trajectories. As AI in robotics continues to grow, it will be crucial to tackle these issues through strong rules and ethical standards to ensure safe and fair progress. Collaborative robots in manufacturing improve safety and increase productivity by working alongside human employees. Autonomous robots reduce human mistakes during checks, leading to better product quality and lower operational expenses. In healthcare, robotic helpers improve patient care and medical staff performance by managing routine tasks. Future research should focus on improving efficiency and accuracy, boosting productivity, and creating safe environments for humans and robots to work safely together. Strong rules and ethical guidelines will be vital for integrating AI-powered robotics into different areas, ensuring technology development aligns with societal values and needs
A dual-model machine learning approach to medicare fraud detection: combining unsupervised anomaly detection with supervised learning
Full text linkMedicare fraud, costing $54.35 billion in improper payments in 2024, undermines U.S. healthcare by draining resources meant for vulnerable populations. Traditional detection methods struggle with reactive designs, high false positives, and reliance on scarce labeled data, exacerbated by a 0.017% fraud prevalence. This paper proposes a dual-model machine learning framework to tackle these challenges. Unsupervised anomaly detection uses cluster-based local outlier factor (CBLOF) and empirical cumulative outlier detection (ECOD) to identify novel fraud patterns across 37 million records. These findings are validated by the list of excluded individuals/entities (LEIE). Supervised classification, with C4.5 decision trees and logistic regression, refines these anomalies using an 80:20 balanced dataset, reducing false positives by 63%. Key innovations include hybrid sampling to address class imbalance, LEIE integration for labeled validation, and parallelized processing of 2.1 million claims hourly. Achieving an area under the curve (AUC), a measure of model accuracy, of 88.3%, this approach outperforms single-model systems by 24%, blending exploratory detection with actionable precision. This scalable, interpretable framework potentially advances fraud detection, safeguarding public funds and Medicare’s integrity with a practical, adaptable solution for evolving threats
Optimizing EfficientNet for imbalanced medical image classification using grey wolf optimization
Full text linkThe advancement of deep learning in computer vision has result in substantial progress, particularly in image classification tasks. However, challenges arise when the model is applied to small and unbalanced datasets, such as X-ray data in medical applications. This study aims to improve the classification performance of fracture X-ray images using the EfficientNet architecture optimized with grey wolf optimization (GWO). EfficientNet was chosen for its efficiency in handling small datasets, while GWO was applied to optimize hyperparameters, including learning rate, weight decay, and dropout to improve model accuracy. Random cropping, rotation, flipping, color jittering, and random erasing, were used to expand the diversity of the dataset, and class weighting is applied to overcome class imbalance. The evaluation uses accuracy, precision, recall, and F1-score metrics. The combination of EfficientNetB0 and GWO resulted in an average 4.5% improvement in model performance over baseline methods. This approach provides benefits in developing deep learning methods for medical image classification, especially in dealing with small and imbalanced datasets
Effects of hyperparameter tuning on random forest regressor in the beef quality prediction model
Full text linkPrediction models for beef meat quality are necessary because production and consumption were significant and increasing yearly. This study aims to create a prediction model for beef freshness quality using the random forest regressor (RFR) algorithm and to improve the accuracy of the predictions using hyperparameter tuning. The use of near-infrared spectroscopy (NIRS) in predicting beef quality is an easy, cheap, and fast technique. This study used six meat quality parameters as prediction target variables for the test. The R² metric was used to evaluate the prediction results and compare the performance of the RFR with default parameters versus the RFR with hyperparameter tuning (RandomSearchCV). Using default parameters, the R-squared (R²) values for color (L*), drip loss (%), pH, storage time (hour), total plate colony (TPC in cfu/g), and water moisture (%) were 0.789, 0.839, 0.734, 0.909, 0.845, and 0.544, respectively. After applying hyperparameter tuning, these R² scores increased to 0.885, 0.931, 0.843, 0.957, 0.903, and 0.739, indicating an overall improvement in the model’s performance. The average performance increase for prediction results for all beef quality parameters is 0.0997 or 14% higher than the default parameters