Computer Engineering and Applications Journal (ComEngApp, Universitas Sriwijaya)
Not a member yet
102 research outputs found
Sort by
Enhanced Short-Term Residential Load Forecasting Using K- means Clustering and Iterative Residual LSTM Networks
No full textAccurate short-term load forecasting (STLF) is essential for optimizing energy management systems, ensuring operational efficiency, and balancing supply and demand in power grids. This study introduces a hybrid model, K-RNLSTM, which integrates K-means clustering with iterative Residual Long Short-Term Memory (LSTM) networks to improve prediction accuracy. The K-means clustering algorithm categorizes similar load patterns, allowing the model to handle seasonal and hourly variations more effectively. Iterative ResBlocks are incorporated within the LSTM framework to capture complex non-linear dependencies and improve the learning process without suffering from degradation. The model was evaluated using real- world residential electricity consumption data across four seasons: winter, spring, summer, and autumn. The K-RNLSTM model consistently outperformed traditional methods such as Extreme Learning Machines (ELM), Seasonal-Trend Loess (STL), Gated Recurrent Units (GRU), and standard LSTM in terms of Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). The results demonstrated that K-RNLSTM achieved an average RMSE of 0.71, MAE of 0.43, and MAPE of 1.31%, surpassing benchmark models across all seasonal variations. Furthermore, the integration of ResBlocks significantly improved the model\u27s ability to minimize large forecasting errors, particularly during peak demand periods. This research demonstrates the effectiveness of combining clustering techniques with deep learning models for short-term load forecasting, offering a robust solution for power system operators to optimize energy distribution and reduce operational costs
Cervical Pre-cancer Classification Using MLP Based on Hybrid Features from GLCM, LBP, and MobileNetV2
No full textThe early and accurate diagnosis of cervical intraepithelial neoplasia lesions (CIN), particularly in a resource-limited environment, is paramount in helping to control the rising epidemic of cervical cancer. This research offers a hybrid classification model that merge texture features like Gray Level Co-occurrence Matrix (GLCM) and Local Binary Pattern (LBP), alongside semantic features from MobileNetV2. These features, after being extracted, are merged and supplied to a Multilayer Perceptron (MLP) for multiclass classification into Normal, CIN1, CIN2, or CIN3. The model was trained and evaluated using a 5-fold stratified cross-validation technique on an IARC dataset that contains 200 cases of colposcopy images. The experimental results illustrate that the model developed with a stratified k-fold cross-validation performed consistently well with high performance, average accuracy reported as 86.75% ± 2.62% and Cohen\u27s kappa 0.7963 ± 0.0524 showed substantial to almost perfect in agreement across folds. The best performance was recorded for Fold 4 achieving 90.31% accuracy, while maintaining robust F1-scores across all classes. This hybrid approach offers a promising direction for developing efficient and accurate computer-aided diagnosis (CAD) systems for cervical lesion classification
Efficient Hierarchical Temporal Audio-Video Cross-Attention Fusion Network For Audio-Enhanced Text-To-Video Retrieval
No full textWith video and audio being integral to modern multimedia content, accurately retrieving relevant segments based on textual queries is crucial for enhancing user experience and information accessibility. However, contextual misalignment across video segments presents significant challenges, particularly when different segments exhibit varying degrees of relevance to specific portions of a text query. To address this issue, a novel Hierarchical Temporal Audio-Video Cross-Attention Fusion Network has been developed. This model utilizes a Video Swim Feature Pyramid video encoder to enhance the extraction of multi-scale spatial features and capture intricate details within videos. Additionally, a Temporal RoBERTa Graph Network serves as the text encoder, enabling a deep understanding of relationships within the text and allowing for minute interpretations of queries that encompass multiple themes. To effectively align video and audio representations with textual queries, the model employs a Hierarchical multiscale spatial-temporal attention mechanism. Furthermore, an Audio Spectrogram Short-Term Memory Transformer is utilized to capture the temporal dynamics of complex audio streams. To refine audio-text alignment, the model incorporates a Threshold-Based audio-text Dynamic Time cross-attention block, which selectively filters irrelevant audio components and dynamically adjusts for temporal misalignments. The experimental results demonstrate that the proposed model significantly enhances retrieval accuracy by effectively aligning video and audio representations with textual queries, resolving multi-scene transitions, and isolating relevant audio cues among complex soundscapes.
 
IoT-Enabled Real-Time Monitoring and Loss-of-Life Estimation of Distribution Transformers
No full textA distribution transformer is required in power distribution networks to step down the voltage relevant and usable for consumers. Its failure not only disrupts electricity supply but also incurs high replacement costs, with broader economic implications. Ensuring reliable operation, therefore, requires accurate and continuous monitoring of its performance. This paper presents IoT-Enabled Real-Time Monitoring and Loss-of-Life Estimation of Distribution Transformers developed and tested on a 10 kVA, 0.415 kV prototype distribution transformer, connected to three residential loads. A dedicated data acquisition system was developed, which monitors key parameters: load current, phase voltage, transformer oil level, ambient temperature, and oil temperature in real time over 14 days. An algorithm was implemented to analyze daily load profiles and hotspot temperature data, which were then used to estimate transformer loss of life. The results show that transformer ageing is highly sensitive to load variation. During weekdays, the cumulative equivalent ageing reached 2.22 hours per day, corresponding to a daily loss of life of 0.00296%. On weekends, higher residential loads increased cumulative ageing to 4.79 hours, with a corresponding life loss of 0.0063%. A simulated one-hour peak load of 1.43 pu resulted in 25.75 hours of ageing, translating to a life loss of 0.034%, demonstrating the severe impact of overloads. These findings emphasize that peak load periods dominate insulation ageing and can substantially reduce service life if unchecked
Analysis and Implementation of Blowfish and LSB Algorithm on RGB Images using SHA-512
No full textThe growth of the internet globally keeps increasing as time goes. There\u27s a big amount of data type saved there too. Those data need to be secured so anyone who doesn\u27t have the right to access them can access it. The purpose of this article is to secure text information into image media using the Blowfish method for encrypting text information and securing it using the Hash function SHA-512 and then embedded it in image media using the Least Significant Bit (LSB) method. The result of implementing those methods using image media sized 138Kb and 39.85Kb with plaintext measuring 27 and 85 characters shows that integrity data is secured with SHA-512 method. The test result using PSNR method to get the score of image quality after embedding information to the image shows that the average number of PSNR’s score is 70,74 dB which means the quality is good and has less difference from the original image
Comparison of Naive Bayes and Support Vector Machine (SVM) Algorithms Regarding The Popularity of Presidential Candidates In The Upcoming 2024 Presidential Election
No full textThis study aims to compare the effectiveness of two classification algorithms, Naive Bayes and Support Vector Machine (SVM), in analyzing the popularity of presidential candidates for the 2024 Presidential Election (Pilpres). The popularity of presidential candidates plays a crucial role in campaign strategies and political decision-making in the modern political era. This research utilizes data from social media, encompassing public sentiment towards presidential candidates and related political issues. The research results indicate that SVM achieves an accuracy rate of 97%, while Naive Bayes achieves 95%, demonstrating the superiority of SVM in predicting the popularity of presidential candidates. In conclusion, the selection of the appropriate algorithm for analyzing complex political data has a significant impact, and the high accuracy rates of both algorithms provide valuable guidance for political decisionmakers and campaign teams in preparation for the upcoming 2024 Pilpres
Turbofan Engine Remaining Useful Life Prediction Using 1-Dimentional Convolutional Neural Network
No full textTurbofan engines have been the dominant type of engine in aircraft for the last forty years. Ensuring the quality of these engines is crucial for flight safety, particularly for long-distance flights. However, their performance degrades over time, impacting flight safety. To address this issue, it is essential to predict potential engine failures by estimating the Remaining Useful Life (RUL) of the engines Deep learning, especially Convolutional Neural Networks (CNNs), has demonstrated exceptional proficiency in handling intricate, non-linear data, leading to improved RUL predictionsdue to their ability to process complex and non-linear data. In this project, a 1-D CNN is used to predict RUL using the NASA C-MAPSS FD001 dataset, which consists of 3 settings and 21 sensors, though sensors with stagnant readings are excluded. The dataset is normalized using min-max and z-score methods, and then segmented into sequences for input into the 1-D CNN model. Various training scenarios were evaluated, with the best RMSE of 3.26 achieved using 10 epochs, a learning rate of 0.0001, and z-score normalization. The results indicate that feature selection can produce a lower RMSE compared to scenarios without feature selection
Cluster Analysis of Obesity Risk Levels Using K-Means And DBScan Methods
No full textObesity is defined as excessive fat accumulation and abnormal accumulation of adipose tissue in the human body that poses health risks. The causes of obesity are multifactorial and include environmental and individual factors. Several factors that cause obesity include genetic, behavioral and environmental factors. Obesity causes various problems in various fields, including health, employment, demographics, economics and family. The problem of obesity has a significant impact on public health. Therefore, understanding and predicting the level of obesity risk is important in efforts to prevent and treat obesity risk. Data on eating habits, physical activity, and other factors associated with obesity levels in certain populations can provide an important basis for understanding obesity risk. This research clusters the risk of obesity to find hidden patterns in the data. The stages in this research consist of pre-processing, clustering, and analysis. The clustering methods used are K-means and DBSCAN. In clustering using the K-means method with a parameter value of k , results are obtained with the same pattern as clustering using the DBSCAN method with a parameter value of epsilon and a minimum sample . In clustering using the K-means method with a parameter value of k , Four clusters were formed which had different patterns. The clustering results obtained in this research can be used as an effort to prevent and treat the risk of obesity
Fake News Detection Using Optimized Convolutional Neural Network and Bidirectional Long Short-Term Memory
No full textThe spread of fake news in the digital age threatens the integrity of online information, influences public opinion, and creates confusion. This study developed and tested a fake news detection model using an enhanced CNN-BiLSTM architecture with GloVe word embedding techniques. The WELFake dataset comprising 72,000 samples was used, with training and testing data ratios of 90:10, 80:20, and 70:30. Preprocessing involved GloVe 100-dimensional word embedding, tokenization, and stopword removal. The CNN-BiLSTM model was optimized with hyperparameter tuning, achieving an accuracy of 96%. A larger training data ratio demonstrated better performance. Results indicate the effectiveness of this model in distinguishing fake news from real news. This study shows that the CNN-BiLSTM architecture with GloVe embedding can achieve high accuracy in fake news detection, with recommendations for further research to explore preprocessing techniques and alternative model architectures for further improvement
Electrical Energy Monitoring and Analysis System At Home Using IoT-Based Prophet Algorithm
No full textElectrical energy is one of the necessities of human life, especially in modern society in urban areas. With a monitoring device for electrical energy consumption using IoT technology, the results of the development show that the monitoring system works well, but the results show that current and voltage measurements are still less accurate. Therefore, in this study, an Electrical Energy Analysis and Monitoring System were developed using the IoT-Based Prophet Algorithm. Data collection was obtained from electrical energy using the PZEM-004T module sensor device used at home and the energy data obtained were stored in a MySQL database. This PZEM data retrieval will appear in real time on the Monitoring Website. The dataset was processed by implementing the Prophet Algorithm, evaluating the model and visualizing the prediction results on the analysis website. Testing using Mean Absolute Percentage Error (MAPE). For design, this system uses energy data and data retrieval time as parameters in the monitoring system for the use of electrical energy at home. Analysis of data taken from electrical energy monitoring was predicted by the model created by the Prophet Algorithm and tested with MAPE to see how accurate the predicted value is in the Prophet Algorithm model. Predictions in this study get an error value of less than 10%, namely 6.87%, which means it is very accurate in predicting the prophet algorithm at home