Bulletin of Electrical Engineering and Informatics
Not a member yet
2885 research outputs found
Sort by
Deep learning-based cellular traffic prediction for 4G long-term evolution networks using three models
Get PDFWireless networks can be seen as the essential element of contemporary communication systems, connecting, in one way or another, billions of people and technologies all over the world. As a result, there is more of requirement from the area of application for models, which should be able to help in the analysis of the time series of mobile traffic to enhance the quality of service (QoS) in the present networks as well as in the future ones. The primary objective of this article is to develop effective artificial intelligence (AI) models for traffic load prediction in cellular networks. To achieve this, we employ three models; gated recurrent unit (GRU), bidirectional long short time memory (BiLSTM), and long short time memory (LSTM), to make numerical estimates of the network traffic at 4G long-term evolution (LTE) cell towers. The empirical results indicate that the BiLSTM model outperforms both the LSTM and GRU models, achieving root mean squared error (RMSE), mean absolute error (MAE), and R2 values of 86.64, 67.12, and 93.23%, respectively. Although this research focuses on traffic modeling for 4G LTE networks, the proposed models hold significant value for the development and optimization of the upcoming generations
Exploring deep learning approaches for image captioning to mimic human understanding
Get PDFImage captioning has emerged as a vital research area in computer vision, aiming to enhance how humans interact with visual content. While progress has been made, challenges like improving caption diversity and accuracy remain. This study proposes transfer learning models and RNN algorithms trained on the microsoft common objects in context (MS COCO) dataset to improve image captioning quality. The models combine image and text features, utilizing ResNet50, VGG16, and InceptionV3 with LSTM, and BiLSTM. Performance is measured using metrics such as BLEU, ROUGE, and METEOR for greedy and beam search. The InceptionV3+BiLSTM model outperformed others, achieving a BLEUscore of over 60%, a METEORscore of 28.6%, and a ROUGEscore of 57.2%. This research contributes to building a simple yet effective image captioning model, providing accurate descriptions with human-like understanding. The error was analyzed to improve results while discussing ongoing research aimed at enhancing the diversity, fluency, and accuracy of generated captions, with significant implications for improving the accessibility and searchability of visual media and informing future research in this area
Image encryption algorithm based on a new one-dimensional chaotic map’s generator
Get PDFEncryption plays a crucial role in protecting sensitive data, including communications, financial transactions, and personal information, from cyber threats. One significant area of encryption is image encryption, which ensures the privacy of visual content, such as in secure image transmission, cloud storage, and medical image processing. Recent advancements in image encryption leverage chaotic maps based on chaos theory, generating unpredictable patterns ideal for securing images. This paper presents a novel chaotic map generator that enhances the dynamics of existing chaotic maps. Based on this generator, we propose a new encryption scheme that operates on the entire input image, obscuring the relationship between the original and encrypted images while spreading pixel changes across the entire encrypted image in one step. The scheme also produces an encrypted image of a different size, making it more efficient and resilient to attacks. While some steps of the proposed system are symmetric, others are asymmetric, ensuring a higher level of security. Based on the obtained results, this approach significantly enhances both security and performance in image encryption
Deep residual bidirectional long short-term memory fusion: achieving superior accuracy in facial emotion recognition
Get PDFFacial emotion recognition (FER) is a crucial task in human communication. Various face emotion recognition models were introduced but often struggle with generalization across different datasets and handling subtle variations in expressions. This study aims to develop the deep residual bidirectional long short-term memory (Bi-LSTM) fusion method to improve FER accuracy. This method combines the strengths of convolutional neural networks (CNN) for spatial feature extraction and Bi-LSTM for capturing temporal dynamics, using residual layers to address the vanishing gradient problem. Testing was performed on three face emotion datasets, and a comparison was made with seventeen models. The results show perfect accuracy on the extended Cohn-Kanade (CK+) and the real-world affective faces database (RAF-DB) datasets and almost perfect accuracy on the face expression recognition plus (FERPlus) dataset. However, the receiver operating characteristic (ROC) curve for the CK+ dataset shows some inconsistencies, indicating potential overfitting. In contrast, the ROC curves for the RAF-DB and FERPlus datasets are consistent with the high accuracy achieved. The proposed method has proven highly efficient and reliable in classifying various facial expressions, making it a robust solution for FER applications
Unsupervised outlier detection in high-dimensional text data: a comparative analysis
Get PDFOutlier detection in user reviews is a critical task for identifying anomalous and potentially valuable insights within large datasets. This study presents a comparative analysis of three different algorithms for outlier detection in user reviews: isolation forest, local outlier factor (LOF), and latent dirichlet allocation (LDA). The performance of each algorithm was evaluated using accuracy and silhouette score for outlier detection and clustering quality. LDA performed best with 0.98 accuracy and a silhouette score of 0.13. Isolation forest followed with 0.90 accuracy and a score of 0.11. LOF had lower results with 0.42 accuracy and a score of -0.05 due to its sensitivity to neighbors. The study contributes by systematically exploring the impact of parameter variations on algorithm performance, providing valuable insights for high-dimensional text data analysis. Despite the promising results, limitations include the dependence on preprocessing and specific parameter settings. Future work will explore hybrid approaches and broader datasets to enhance scalability and adaptability
Medication box management system with automatic dosing integrated with IoT-based Android app and Firebase
Get PDFUtilizing Firebase technology and an Android internet of things (IoT) application, the research endeavors to create a smart medicine box in order to enhance the efficacy of automated drug management. Hardware implementation, software implementation, and 3D design planning for automatic dosage adjustment are the methods utilized. The results prove that the application effectively controls the dosage, evacuation schedule, and quantity of the medication based on the user’s input. Boundary value analysis (BVA) black box testing demonstrated that every feature of the application functions as intended. Furthermore, the efficacy of drug production testing indicates that the smart medicine box exhibits a notable level of precision, albeit with a limited number of inaccuracies that could be rectifiable through additional parameter and mechanism optimizations within the drug box. Consequently, the investigation has effectively produced an automated drug management system that has the potential to enhance drug use supervision and safety, particularly for elderly services individuals residing alone
Predicting demand in changing environments: a review on the use of reinforcement learning in forecasting models
Get PDFThis systematic review, carried out under the PRISMA methodology, aims to identify how reinforcement learning has been used in demand forecasting, distinguishing the problems they are trying to overcome, recognizing the algorithms used, detailing the performance metrics used, recognizing the performance achieved by these models and identifying the business sectors in which it has been developed. Studies from all sectors were considered to expand the search range. A total of 24 articles were qualitatively analyzed, and the main results were that reinforcement learning has been used mainly for the selection or dynamic integration of the best predictors from a base of them to adapt to changing environments; whereas forecasting in volatile and complex environments is the main issue addressed; whereas Q-learning (QL), deep q network (DQN), double deep q network (DDQN), and deep deterministic policy gradient (DDPG) are the most widely used algorithms; and that, finally, the sectors of electric power, thermal energy, transport and telecommunications are the sectors where this type of forecast has been developed. Finally, given that all the models studied lack mechanisms for detecting concept drift, a new use of reinforcement learning for this purpose is proposed
An efficient course recommendation system for higher education students using machine learning techniques
Get PDFEducation institutions and teachers are in desperate need of automated, non-intrusive means of getting student feedback that would allow them to better understand the learning cycle and assess the success of course design. Students would benefit from a framework that intelligently guides their actions and provides exercises or resources to support and enhance their learning. The recommender system framework is a software agent that learns the user's preferences through a variety of channels and then utilizes that knowledge to provide product suggestions. A recommendation engine considers all potential user interests as background information, uses that knowledge to produce convincing recommendations, and then returns those ideas to the user. This article presents a feature selection and machine learning based course recommendation system for higher education students. principal component analysis (PCA) algorithm is used for feature selection. AdaBoost, k nearest neighbour (KNN), and Naïve Bayes algorithms are used to classify and predict student data. It is found that the AdaBoost algorithm is having better accuracy and F1 score for course recommendation to students. PCA AdaBoost is achieving an accuracy of 99.5%
Strategic processor task allocation through game-theoretic modeling in distributed computing environments
Get PDFThis paper explores a game-theoretic model for task allocation in distributed systems, where processors with varying speeds and external load factors are considered strategic players. The goal is to understand the impact of processors' strategic behaviors on workload management and overall system efficiency, focusing on the attainment of a pure strategy Nash Equilibrium (NE). The research rigorously develops a formal mathematical model and validates it through extensive simulations, highlighting how NE ensures stability but may not always yield optimal system performance. The adaptive algorithms for dynamic task allocation are proposed to enhance efficiency in real-time processing environments. Results demonstrate that while NE provides stability, the adoption of optimal cooperative strategies significantly improves operational efficiency and reduces transaction costs. The findings contribute valuable insights into the strategic interactions within computational frameworks, offering guidelines for developing more efficient systems. This study not only advances the theoretical understanding of strategic task allocation but also has practical implications for system design and policy-making in areas such as cloud computing and traffic management
Optimized colon cancer classification via feature selection and machine learning
Get PDFThe increasing dimensionality of gene expression data poses significant challenges in cancer classification, particularly in colon cancer. This study presents a novel filtering approach (FA) and a gene classifier (GC) to enhance gene selection and classification accuracy. Utilizing a dataset of 62 samples, our methods integrate statistical measures and machine learning classifiers, achieving classification accuracies of 96% and 97%, respectively. The FA effectively filters out noise and redundancy, allowing for accurate predictions with a minimal subset of genes, while the GC leverages multiple classifiers for optimal performance. These findings underscore the importance of robust feature selection in improving cancer diagnostics and suggest potential applications in personalized medicine. By addressing the limitations of existing methodologies, our work lays the groundwork for future research in cancer genomics, emphasizing the need for adaptive strategies to handle complex datasets