Irish Studies Review

Artificial electric field algorithm (AEFA) is a metaheuristic optimization algorithm proposed in recent years, which has been successfully applied to address various optimization problems. However, it is likely to converge prematurely or fall into local optima when solving complex problems. To overcome these disadvantages, a multi-strategy artificial electric field algorithm (MAEFA) is proposed in this paper. For the MAEFA algorithm, the global optimal solution information is utilized to improve the diversity of population and global search ability. Then, the adaptive Coulomb’s constant is configured to balance the global exploration and local search. Also, a restart strategy is designed to further alleviate the premature convergence. To validate the effectiveness of MAEFA, it is compared with three AEFA algorithms and several other evolutionary algorithms on 14 test problems presented in CEC 2005 and 13 basic benchmark functions. Furthermore, a wind power prediction model based on MAEFA algorithm and back-propagation (BP) neural network is established to investigate its application ability. Experiments show that MAEFA is significantly superior to other algorithms in tackling these benchmark functions with different dimensions. Furthermore, in terms of wind power prediction, the BP neural network model optimized by MAEFA algorithm also provides higher prediction accuracy.

Nowadays, the Deepfake technology is mainly used to harm people’s reputations and can trick the face recognition system by swapping faces between people, raising significant security concerns. Thus, methods for detecting Deepfake are crucial. The recent methods for Deepfake detection have performed well in distinguishing real content from fake content. Some research employed the Transformer technique, commonly used in natural language processing (NLP), to enhance performance. Therefore, this paper proposes a novel deepfake detection method that transforms extracted features into words and utilizes NLP techniques for deepfake classification. We employed a fine-tuned pre-trained Convolutional Neural Network (CNN) model to extract features from the face images in the videos. These extracted features are labeled based on grouping methods, such as mean and standard deviation (SD). Tokenization and classification are then performed using Long Short-Term Memory (LSTM) and Recurrent Neural Network (RNN). Additionally, Bidirectional Encoder Representations from Transformers (BERT) is used as another tokenizer and classifier to compare the performance of deepfake detection between the traditional model and the NLP model. The result states that the method using BERT as a tokenizer and classifier with Mean and SD grouping method shows better efficiency, achieving 99.57% on the Roc Curve, 99.58% Accuracy, 99.18% Precision, 100.00% recall, and 99.59% F-measure.

Personalized recommendation systems play a crucial role in alleviating information overload and satisfying users’ specific preferences. To address the challenges of inadequate user historical data extraction and the cold start problem inherent in traditional movie recommendation systems, we present a novel personalized movie recommendation model known as “movie recommendation with starring roles and ratings” (MSR). By incorporating a multi-head attention mechanism, the model captures intricate relationships among diverse data fields within users’ viewing records and facilitates the extraction of user features through the basic information-rating joint attention network (BRJA). The gate mechanism efficiently integrates fundamental movie information and average score into the movie representation vector, thereby generating candidate movie features. MSR can effectively provide recommendations even when confronted with limited user information, effectively mitigating the cold start problem. Comparative experiments on the movie lens dataset and ablation experiments focusing on key modules demonstrate the effectiveness of MSR in improving movie recommendations.

Multi-dimensional Flight Trajectory Prediction (MFTP) in Flight Operations Quality Assessment (FOQA) refers to the estimation of flight status at the future time, accurate prediction future flight positions, flight attitude and aero-engine monitoring parameters are its goals. Due to differences between flight trajectories and other kinds trajectories and difficult access to data and complex domain knowledge, MFTP in FOQA is much more challenging than Flight Trajectory Prediction (FTP) in Air Traffic Control (ATC) and other trajectory prediction. In this work, a deep Koopman neural operator-based multi-dimensional flight trajectory prediction framework, called Deep Koopman Neural Operator-Based Multi-Dimensional Flight Trajectories Prediction (FlightKoopman), is first proposed to address this challenge. This framework is based on data-driven Koopman theory, enables to construct a prediction model using only data without any prior knowledge, and approximate operator pattern to capture flight maneuver for downstream tasks. The framework recovers the complete state space of the flight dynamics system with Hankle embedding and reconstructs its phase space, and combines a fully connected neural network to generate the observation function of the state space and the approximation matrix of the Koopman operator to obtain an overall model for predicting the evolution. The paper also reveals a virgin dataset Civil Aviation Flight University of China (CAFUC) that could be used for MFTP tasks or other flight trajectory tasks. CAFUC Datasets and code is available at this repository: https://github.com/CAFUC-JJJ/FlightKoopman . Experiments on the real-world dataset demonstrate that FlightKoopman outperforms other baselines.

Autism spectrum disorder (ASD) is a developmental disability that poses significant challenges in social interaction, communication, and behavior. Individuals with ASD have unique ways of interacting and communicating, and early prediction is crucial for timely therapy. Researchers are focusing on predicting ASD using image-processing techniques due to its neurological nature. The proposed novel Hybrid Convolutional Bilateral filter-based Deep Dual Swin Axial Generator Attention with Gooseneck Barnacle Optimization (FCB-DDSATGA-GBO) accurately predicts ASD. The facial image dataset is the input data source. The Hybrid Fast Convolutional Bilateral Filter (HFCBF) is used to pre-process the data. Dual Deep Autoencoder and Split Generative Adversarial Network (DDASGAN) is used to extract static features. Additionally, Swin-Gated Axial Attention Transformer (SGAAT) is used to segment the image. To forecast ASD, DDASGAN is used and optimized with Gooseneck Barnacle Optimization (GBO). The performance of the suggested methodology can be assessed using measures such as accuracy, recall, precision, sensitivity, f-score, and error, and compared to existing methods. The suggested FCB-DDSATGA-GBO model outperforms the current techniques, offering an enhanced f1-score of 99.66%, recall of 99.66%, accuracy of 99.67%, specificity of 99.67%, and precision of 99.66% when utilizing facial images.

The integration of solar photovoltaic (SPV) systems with modular multiport converters (MMPC) enables efficient energy conversion and distribution, enhancing the overall performance and reliability of renewable energy systems (RES). However, the complexity of the control algorithms and potential issues related to the dynamic response can pose challenges in achieving optimal performance and stability in varying operating conditions. This paper proposes a hybrid method for integrating SPV systems with MMPC to achieve efficient power management in modern renewable energy grids. The proposed hybrid method is the combined execution of the Osprey Optimization Algorithm (OOA) and Relational Bi-level Aggregation Graph Convolutional Network (RBAGCN). Hence it is named as OOA-RBAGCN technique. The aim is to ensure optimal power transfer, minimize total harmonic distortion (THD), maintain voltage stability under dynamic operating conditions, and ultimately improve the overall energy efficiency, reliability, and performance of SPV-based RES within smart grid applications. The OOA is used to optimize the control parameter of the proportional-integral (PI) controller. The RBAGCN is used to predict these optimized parameters. By then, the proposed approach is used on the MATLAB platform and compared with other approaches such as Starling Murmuration Optimization (SMO), Dung Beetle Optimizer (DBO), Improved Harris Hawks Optimization (IHHO), Grey Wolf Optimization (GWO), and Particle Swarm Optimization (PSO). The proposed method achieves a high efficiency of 98.1%, and a reduced THD of 2.9% significantly surpassing all existing methods.

The rising incidence of heart disease requires effective and robust prediction algorithms, especially in Internet of Things (IoT)-cloud-based smart healthcare frameworks. This study presents a novel method for forecasting cardiovascular disease using superior data preprocessing, feature selection, and deep learning techniques. First, preprocessing is done using the Z-score min–max normalization technique to ensure consistent data scaling and standardize the dataset. After preprocessing, an innovative hybrid feature selection technique that combines Black Widow Optimization (BWO) and Influencer Buddy Optimization (IBO) is utilized. By achieving equilibrium between invention and execution, the BWO-IBO technique enhances feature selection and extracts the most pertinent information for heart disease prediction. The Gates-Controlled Deep Unfolding Network (GCDUN), which is based on the Crayfish Optimization Algorithm (COA), is an innovative framework for prediction. Through the use of a gates-controlled mechanism and a COA component that speeds up network parameter tuning based on crayfish behavior, GCDUN-COA increases feature representation and enhances the decision plane. The fusion of the IoT and a cloud-based framework takes the present data collection, processing, and remote monitoring a notch higher, thus making the system highly scalable and efficient for clinical use. When predicting cardiac disease, the method recommended shows improved F1-score, specificity, accuracy, recall, and precision continuously achieving above 99% across all performance metrics. By providing prompt diagnosis and intervention via an intelligent, adaptive prediction system, an IoT-driven cloud-based medical technology has the potential to revolutionize cardiac care.

Knowledge-based visual question answering relies on open-ended external knowledge and a fine-grained comprehension of both the visual content of images and semantic information. Existing methods for utilizing knowledge have the following limitations: (1) Language pre-training methods output answers in the form of plain text, which only understand shallow visual content; (2) The knowledge retrieved by image objects as labels is represented as first-order logic, making it difficult to infer complex questions. To address the above problems, this paper integrates visual-textual multimodal information, accumulates domain-specific and external multi-modal knowledge, introduces and supplements external objective facts, and proposes a multimodal knowledge graph construction and fact-assisted reasoning network (MKGFA). The network consists of three parts: the multimodal knowledge graph construction module (MKGC), the objective fact-assisted reasoning module (FAR), and the answer inference module. The MKGC engages in the coarse-to-fine-grained learning of triplet representations for multimodal knowledge units. The FAR establishes deep cross-modal relations between visual objects and factual words for correlating real answers. The answer inference module makes the final decision based on the results of both. Among them, the former two modules employ a pre-training and fine-tuning strategy, systematically accumulating foundational and domain-specific knowledge. Compared with the state-of-the-arts, MKGFA achieves 1.09% and 0.7% higher accuracy on the two challenging OKVQA and KRVQA datasets, respectively. The experimental results demonstrate the complementary advantages of the integration of the two modules.

With the development of machine vision and multimedia technology, posture detection and related algorithms have become widely used in the field of human posture recognition. Traditional video surveillance methods have the disadvantages of slow detection speed, low accuracy, interference from occlusions, and poor real-time performance. This paper proposes a real-time pose detection algorithm based on deep learning, which can effectively perform real-time tracking and detection of single and multiple individuals in different indoor and outdoor environments and at different distances. First, a corresponding pose recognition dataset for complex scenes was created based on the YOLO network. Then, the OpenPose method was used to detect key points of the human body. Finally, the Kalman filter multi-object tracking method was used to predict the state of human targets within the occluded area. Real-time detection of human postures (sitting, stand up, standing, sit down, walking, fall down, and lying down) is achieved with corresponding alarms to ensure the timely detection and processing of emergencies.

Biotechnological analysis of DNA microarray genes provides valuable insights into the discovery and treatment of diseases such as cancer. It may also be crucial for the prevention and treatment of other genetic diseases. However, due to the large number of features and dimensions in a DNA microarray, the “curse of dimensions” problem is very common. Many machine learning methods require an effective subset of input genes to achieve high accuracy. Unfortunately, extracting features (genes) is an inherently NP-hard problem. Recently, the use of metaheuristics to overcome the NP-hardness of the feature extraction problem has attracted the attention of many researchers. In this paper, we use the combination of fuzzy entropy and Giza Pyramid Construction (GPC) for feature selection. First, redundant features in the microarray dataset are removed using the fuzzy entropy approach. GPC is then used to reduce the execution time. This results in the selection of a near-optimal subset of genes for cancer detection. Dimensionality reduction with GPC followed by classification with Convolutional Neural Network (CNN) creates a synergy to increase efficiency. The proposed method is tested on five well-known cancer patient datasets: leukemia, lymphoma, MLL, ovarian, and SRBCT. The performance of CNN was also measured with four well-known classifiers, including K-nearest neighbor, naïve Bayesian, decision tree, and logistic regression. Our results show that, on average, CNN has the highest accuracy, recall, precision, and F-measure in all datasets.

In the medical field, Internet of Things (IoT) applications allow for real-time diagnosis and remote patient monitoring, commonly called Internet of Health Things (IoHT). However, cybersecurity attacks may interrupt hospital operations and threaten patients’ health and well-being due to this integration. Hence, developing an Intrusion Detection System (IDS) suited explicitly for healthcare systems is essential to ensure efficiency and accuracy. Nevertheless, it is challenging to integrate anomaly-based IDS frameworks in healthcare systems as they necessitate additional processing time, temporal feature retention, and increased complexity. Therefore, a deep learning system based on SqueezeNet and NasNet is presented in this paper to detect intrusions in a healthcare setting. In this, SqueezeNet is employed to extract more significant features. On the other hand, network breaches while data transmission across distinct locations are detected by the NasNet-based classifier. In addition, the Rider Optimization Algorithm (ROA) is applied to adjust the classifier’s hyperparameters, guaranteeing that it would accurately detect attacks. Moreover, the Auxiliary Classifier Generative Adversarial Network (ACGAN) approach is integrated into the proposed framework to avoid data imbalance. Applying different performance constraints, the proposed approach is thoroughly assessed on three publicly available datasets (TON-IoT, ECU-IoHT, and WUSTL-EHMS). The results show that the proposed deep learning-based cybersecurity model outperforms traditional methods and produces better outcomes.

Generative adversarial networks (GANs) have been widely applied in the field of meteorological research, particularly in the downscaling of images due to their ability to generate super-resolution images. In recent years, numerous researchers have combined GANs with recurrent neural networks (RNNs) to address the issue of meteorological super-resolution. However, these models do not take into account the spatial variations of meteorological sequences. In this paper, we propose a super-resolution method named RN-STLSTM-GAN, which combines GANs with RN-STLSTM and ESA networks to learn the spatiotemporal features of meteorological sequences. Specifically, we first apply the RN-STLSTM at the initialization of the generator and discriminator to learn the spatiotemporal relationships between sequential images. Second, an ESA network is combined with the RN-STLSTM structure to enhance the learning of spatial features. Thirdly, LeakyReLU is used as the activation function for both the generator and discriminator to minimize the loss of image data during model training. Experiments conducted on the NJU-CPOL datasets demonstrate that our proposed method outperforms other existing methods and can generate realistic and temporally consistent super-resolution sequences for datasets at different heights.

The objective of image captioning is to empower computers to generate human-like sentences autonomously, describing a provided image. To tackle the challenges of insufficient accuracy in image feature extraction and underutilization of visual information, we present a Swin Transformer-based model for image captioning with feature enhancement and multi-stage fusion (Swin-Caption). Initially, the Swin Transformer is employed in the capacity of an encoder for extracting images, while feature enhancement is adopted to gather additional image feature information. Subsequently, a multi-stage image and semantic fusion module is constructed to utilize the semantic information from past time steps. Lastly, a two-layer LSTM is utilized to decode semantic and image data, generating captions. The proposed model outperforms the baseline model in experimental tests and instance analysis on the public datasets Flickr8K, Flickr30K, and MS-COCO.