Intelligent Systems with Applications最新文献

Optimal reactive power dispatch (ORPD) problems are important tools for the sake of security and economics of power systems. The ORPD problems are nonlinear optimization problems to minimize the real power losses and voltage profile enhancement by optimizing several discrete and continuous control variables. This paper proposes a Lévy-flight phasor particle swarm optimization (LPPSO) for solving ORPD problems while considering real power losses and voltage profile in two standard power systems. The simulation results demonstrate that the LPPSO algorithm proves itself as an acceptable method for reaching a more optimal solution for the ORPD problems.

Internet of Things (IoT) devices are extensively utilized but are susceptible to cyberattacks, posing significant security challenges. To mitigate these threats, machine learning techniques have been implemented for network intrusion detection in IoT environments. These techniques commonly employ various feature reduction methods, prior to inputting data into models, in order to enhance the efficiency of detection processes to meet real-time requirements. This study provides a comprehensive comparison of feature selection (FS) and feature extraction (FE) techniques for network intrusion detection systems (NIDS) in IoT environments, utilizing the TON-IoT and BoT-IoT datasets for both binary and multi-class classification tasks. We evaluated FS methods, including Pearson correlation and Chi-square, and FE methods, such as Principal Component Analysis (PCA) and Autoencoders (AE), across five classic machine learning models: Decision Tree (DT), Random Forest (RF), Naive Bayes (NB), k-Nearest Neighbors (kNN), and Multi-Layer Perceptron (MLP). Our analysis revealed that FE techniques generally achieve higher accuracy and robustness compared to FS methods, with RF paired with AE delivering superior performance despite higher computational demands. DTs are most effective with smaller feature sets, while MLPs excel with larger sets. Chi-square is identified as the most efficient FS method, balancing performance and computational efficiency, whereas PCA outperforms AE in runtime efficiency. The study also highlights that FE methods are more effective for complex datasets and less sensitive to feature set size, whereas FS methods show significant performance improvements with more informative features. Despite the higher computational costs of FE methods, they demonstrate a greater capability to detect diverse attack types, making them particularly suitable for complex IoT environments. These findings are crucial for both academic research and industry applications, providing insights into optimizing detection performance and computational efficiency in NIDS for IoT networks.

Attention is a very popular and effective mechanism in artificial neural network-based sequence-to-sequence models. In this survey paper, a comprehensive review of the different attention models used in developing automatic speech recognition systems is provided. The paper focuses on how attention models have grown and changed for offline and streaming speech recognition in recurrent neural networks and Transformer-based systems.

The underground water pipeline system is a crucial infrastructure that largely remains out of sight. However, it is the source of a clean and uninterrupted flow of water for our everyday lives. Various factors, including corrosion, material degradation, ground movement, and improper maintenance, cause pipe leaks, a silent crisis that causes an estimated 39 billion dollars of loss every year. Prompt leakage detection and localization can help reduce the loss. This research investigates the potential of two machine learning models as supporting tools for surveying extensive areas to identify and pinpoint the location of underground leaks. The presented combined approach ensures the speed and accuracy of the leakage survey. The first machine learning model is a hybrid ML model that employs thermal imaging to identify subterranean water leakage. It relies on detecting thermal anomalies and distinctive signatures associated with water leakage to identify and locate underground water leakage. The developed model can detect up to 750 mm underground leakage with 95.20 % accuracy. The second model uses binaural audio from geophones to localize the leakage position. The model utilizes interaural time difference and interaural phase difference for localization purposes, and the 1D-CNN network delivers an angle in twenty-degree increments with an accuracy of 88.19 %. Large-scale implementation of the proposed model could be a powerful catalyst to reduce water loss in the water supply system.

This paper introduces a novel approach to personal risk (PR) identification using federated learning (FL) in wireless communication scenarios, leveraging generalized information. The primary focus is on harnessing the power of distributed data across various wireless devices while ensuring data privacy and security, a critical concern in PR assessment. To this end, we propose an FL-based model that effectively aggregates learning from diverse, decentralized data sources to analyze the PR factors. The proposed method involves training local models on individual devices, which are then aggregated to form a comprehensive global model. This process not only preserves data privacy by keeping sensitive information on the device but also utilizes the widespread availability and connectivity of wireless devices to enhance data richness and model robustness. To address the challenges posed by the wireless environment, such as data heterogeneity and communication constraints, we further implement advanced aggregation algorithms and optimization techniques tailored to these unique conditions. We finally evaluate the performance of our proposed method based on two primary metrics of identification accuracy and convergence rate of the federated learning process. Through extensive simulations and real-world experiments, we demonstrate that our approach not only achieves high accuracy in PR identification but also ensures rapid convergence, making it a viable solution for real-time risk assessment in wireless networks.

This paper introduces a novel approach utilizing normalized flow networks (NFNs) for dynamic personal risk (PR) analysis, specifically focusing on the assessment of two-way data rates at network nodes. NFNs, a sophisticated paradigm in data processing and modeling derived from machine learning principles, serve as the foundational framework for our analysis. Leveraging NFNs, we develop a generalized method that integrates information transmission techniques into PR dynamics, enabling a comprehensive examination of communication efficacy within network structures. Our study entails the formulation of dynamic models tailored to capture the evolving nature of PR interactions, facilitating the evaluation of data rates exchanged between network nodes. Through extensive simulations and empirical validation, we demonstrate the effectiveness of our approach in elucidating the intricate dynamics of PR campaigns and quantifying the impact on the network performance. The findings underscore the significance of leveraging NFNs for dynamic PR analysis, offering valuable insights into optimizing communication strategies and enhancing network efficiency in diverse domains.

Ribonucleic acid is a crucial biomolecule in living organisms, with various types. To promote the research process of ribonucleic acid function, this study is for analyzing the utilization of heuristic algorithms on the ground of fusion dynamic programming in predicting the secondary structure of ribonucleic acid. Research on novel use of tree models for RNA secondary structure comparison, and use heuristic algorithms to optimize the multi branch structure comparison of tree models. Firstly, this study utilized dynamic programming algorithms to construct a comparison matrix and successfully found the backtracking path in the matrix. Meanwhile, for ensuring that the structural information of ribonucleic acid is not lost during the comparative analysis process, the study applies the idea of heuristic algorithms to calculate the optimal comparison between multi branched loops. Finally, the weights are adjusted using neural network algorithms to predict the optimal alignment structure. The results showed that the fusion dynamic programming heuristic algorithm achieved generalization performance of 0.928, 0.856, 0.842, and 0.793 on the target base data test sets of humans, mice, yeast, and spotted fish, respectively. Compared with the SimTree algorithm, the generalization performance has been improved by 15.13 %, 27.38 %, 27.77 %, and 38.88 %, respectively. In summary, the application of heuristic algorithms integrating dynamic programming in predicting the secondary structure of ribonucleic acid has good predictive performance. This has reference value for a deeper understanding of the structure and function relationship of ribonucleic acid.

Chronic Kidney Disease (CKD) is increasingly recognised as a major health concern due to its rising prevalence. The average survival period without functioning kidneys is typically limited to approximately 18 days, creating a significant need for kidney transplants and dialysis. Early detection of CKD is crucial, and machine learning methods have proven effective in diagnosing the condition, despite their often opaque decision-making processes. This study utilised explainable machine learning to predict CKD, thereby overcoming the 'black box' nature of traditional machine learning predictions. Of the six machine learning algorithms evaluated, the extreme gradient boost (XGB) demonstrated the highest accuracy. For interpretability, the study employed Shapley Additive Explanations (SHAP) and Partial Dependency Plots (PDP), which elucidate the rationale behind the predictions and support the decision-making process. Moreover, for the first time, a graphical user interface with explanations was developed to diagnose the likelihood of CKD. Given the critical nature and high stakes of CKD, the use of explainable machine learning can aid healthcare professionals in making accurate diagnoses and identifying root causes.

Clinicians benefit from the use of artificial intelligence and machine learning techniques applied to health data within health records, which identify commonalities between them. It enables them to get evidence-based support in recommending shared treatment paths for undiagnosed health records. The collective inference from these patterns, drawn from an array of health records, further enhances the capacity to mine essential features, supporting public health experts in their management of population health conditions. This paper presents a novel mapping tool model designed to analyze electronic health records and provide healthcare providers with evidence-based decision support. The work focuses on the analysis of health records from hospital databases, encompassing parameters extracted from routine health checkups. By scrutinizing patterns within examined health records, healthcare providers can seamlessly align with newer health records for diagnosis and treatment recommendations. Core to this approach is the integration of a fuzzy rule-based classifier system within the proposed system. This incorporation facilitates the processing of health records, extracting pertinent features to augment decision-making with the support of knowledge bases. The model architecture provides flexibility and customizability, enabling easy configuration of the system to accurately map new health records to the examined dataset. Additionally, the model utilizes a specially developed distance-measure technique tailored for the proposed fuzzy-based system. Results showcase satisfying performance and robust discriminant capability for accurate recommendations. The alignment of outcomes with expert evaluations underscores the model's efficacy and attainment of benchmarks.

The COVID-19 virus is causing a global pandemic. The total number of new coronavirus cases worldwide by the end of November 2020 had already surpassed 60 million. The World Health Organization (WHO) has determined that wearing masks is a crucial precaution during the COVID-19 epidemic to limit the growth of viruses, and facemasks are frequently seen in public places worldwide. Also, many public service providers wear face masks (covering their mouths and noses). These events brought attention to the need for automatic computer-vision-based object detection (masked face detection) methods to track public behavior. Therefore, it is necessary to develop tools for monitor people who have not used masks in public service areas in real-time. Reducing the spread of infectious diseases can occur when masked face detection techniques are used for authentication instead of mask removal for face matching. A superior framework of masked face detection could improve security systems and lower the rate of crime. Masked face detection is a computer vision method standard in people's daily lives to recognize, discover, and recognize masked faces in pictures and videos. This study provides a thorough and systematic analysis of masked face detection algorithms. With the help of examples, we have thoroughly examined and reviewed the studies done concerning face mask identification and techniques for masked face detection.

Additionally, we compared and explained different masked face detection dataset types, libraries, and techniques. We also discussed the challenges with masked face detection and whether the researchers could overcome them. We have discussed and conducted a thorough evaluation of the accuracy, pros, and cons of various approaches by comparing their performance on multiple datasets. As a result, this study aims to give the researcher a broader viewpoint to aid him in finding patterns and trends in masked face detection in various COVID-19 contexts, overcoming challenges that are still present, and creating future algorithms for masked face detection that are more reliable and accurate.