Search

Electroencephalography (EEG) serves as a crucial neurodiagnostic tool by recording the electrical brain activity via attached electrodes on the patient's head. While artificial intelligence (AI) exhibited considerable promise in medical diagnostics, its potential in the realm of neurodiagnostics remains underexplored. This research addresses this gap by proposing an innovative approach employing time-series classification of EEG data, leveraging long-short-term memory (LSTM) neural networks for the identification of abnormal brain activity, particularly seizures. To enhance the performance of the proposed model, metaheuristic algorithms were employed for optimizing hyperparameter collection. Additionally, a tailored modification of the variable neighborhood search (VNS) is introduced, specifically tailored for this neurodiagnostic application. The effectiveness of this methodology is evaluated using a carefully curated dataset comprising real-world EEG recordings from both healthy individuals and those affected by epilepsy. This software-based approach demonstrates noteworthy results, showcasing its efficacy in anomaly and seizure detection, even when working with relatively modest sample sizes. This research contributes to the field by illuminating the potential of AI in neurodiagnostics, presenting a methodology that enhances accuracy in identifying abnormal brain activities, with implications for improved patient care and diagnostic precision. [ABSTRACT FROM AUTHOR]

In recent years, lithium-ion batteries (LIB) have become the de facto energy storage means for electric vehicles (EVs) due to their high energy density. However, LIBs require state of charge (SOC) monitoring to ensure safe operating conditions and for an enhanced lifespan. Since SOC cannot be directly measured, various estimation methods have been proposed in recent literature, most notably the recent rise in popularity of long short-term memory-recurrent neural networks (LSTM-RNN). Current research in the use of LSTM-RNNs typically applies a single strong data-driven model that can produce accurate predictions at the expense of lengthy model training times. As LSTM-RNNs must be retrained as LIBs age to maintain reasonable estimation accuracies, this poses a problem for EV battery management system processors. To address this research gap, this work proposes a homogeneous ensemble learning model based on several LSTM-RNN base models, as a solution to reduce the training time. The LSTM base models are fused by a meta-learner, to overcome the shortcomings of traditional ensemble fusion methods. Data diversification methods for homogeneous ensembles are also reviewed and benchmarked in this paper. The proposed method achieves a low model training time by 2.6–3.5 times while maintaining a similar mean absolute error (MAE) of 1.4% when compared to conventional shallow and deep LSTM-RNN models. The proposed model was also successfully validated with battery discharge data collected from a custom build battery tester. It is anticipated that the proposed LIB SOC estimation approach can contribute to increased feasibility of using artificial intelligence in EVs in general and improve EV battery management. [ABSTRACT FROM AUTHOR]

Lithium-ion batteries' state-of-charge prediction (SoC) cannot be directly measured due to their chemical structure. Therefore, a prediction can be made using the measurable data of the battery. The limited measurable data (current, voltage, and temperature) and the small changes in charge/discharge curves over time further complicate the prediction process. Recurrent neural network-based deep learning algorithms, capable of making predictions with a small number of input data, have become widely used in this field. Particularly, the use of Long Short-Term Memory (LSTM) has shown successful results in one-dimensional and slowly changing data over time. However, these approaches require high computational power for training and testing processes. The window length of the data used as input is one of the major factors affecting the prediction time. The window length of the data varies depending on the sampling frequency and the length of the lookback period. Reducing the window length to shorten, the prediction time makes feature extraction from the data difficult. In this case, adjusting the sampling frequency and window length properly will improve the prediction accuracy and time. Therefore, this study presents the effects of sampling frequency and window length on the prediction accuracy for LSTM-based deep learning approaches. Prediction results were examined using different metrics such as MAE, MSE, training, and testing time. The study's results indicate that training and testing times can be shortened when the sampling frequency and window length are properly adjusted. [ABSTRACT FROM AUTHOR]

The accuracy of long-term runoff models can be increased through the input of local weather variables and global climate indices. However, existing methods do not effectively extract important information from complex input factors across various temporal and spatial dimensions, thereby contributing to inaccurate predictions of long-term runoff. In this study, local–global–temporal attention mechanisms (LGTA) were proposed for capturing crucial information on global climate indices on monthly, annual, and interannual time scales. The graph attention network (GAT) was employed to extract geographical topological information of meteorological stations, based on remotely sensed elevation data. A long-term runoff prediction model was established based on long-short-term memory (LSTM) integrated with GAT and LGTA, referred to as GAT–LGTA–LSTM. The proposed model was compared to five comparative models (LGTA–LSTM, GAT–GTA–LSTM, GTA–LSTM, GAT–GA–LSTM, GA–LSTM). The models were applied to forecast the long-term runoff at Luning and Pingshan stations in China. The results indicated that the GAT–LGTA–LSTM model demonstrated the best forecasting performance among the comparative models. The Nash–Sutcliffe Efficiency (NSE) of GAT–LGTA–LSTM at the Luning and Pingshan stations reached 0.87 and 0.89, respectively. Compared to the GA–LSTM benchmark model, the GAT–LGTA–LSTM model demonstrated an average increase in NSE of 0.07, an average increase in Kling–Gupta Efficiency (KGE) of 0.08, and an average reduction in mean absolute percent error (MAPE) of 0.12. The excellent performance of the proposed model is attributed to the following: (1) local attention mechanism assigns a higher weight to key global climate indices at a monthly scale, enhancing the ability of global and temporal attention mechanisms to capture the critical information at annual and interannual scales and (2) the global attention mechanism integrated with GAT effectively extracts crucial temporal and spatial information from precipitation and remotely-sensed elevation data. Furthermore, attention visualization reveals that various global climate indices contribute differently to runoff predictions across distinct months. The global climate indices corresponding to specific seasons or months should be selected to forecast the respective monthly runoff. [ABSTRACT FROM AUTHOR]

In the rapidly evolving landscape of network communication systems, the need for robust security measures has become paramount due to increased vulnerability to cyber threats. Traditional Intrusion Detection Systems (IDSs) face challenges in efficiently handling redundant features, leading to increased computational complexity. This research addresses these challenges by proposing two optimized IDSs leveraging Grey Wolf Optimization (GWO) combined with deep learning (DL) models. The first system integrates Gated Recurrent Unit (GRU) with GWO (GRU-GWO), while the second utilizes Long Short-Term Memory (LSTM) with GWO (LSTM-GWO). These systems aim to enhance feature selection, reducing dimensionality and improving detection accuracy. The NSL-KDD and UNSW-NB15 datasets, representative of contemporary network environments, were employed to evaluate the proposed systems. Experimental results demonstrate significant improvements in intrusion detection accuracy and computational efficiency, underscoring the efficacy of the DL-GWO approach in enhancing network security. The first approach (GRU-GWO-FS) increased accuracy to 90% and 79% for anomaly and signature-based detection on the UNSW-NB15 dataset, compared to 80% and 77% with all features. The second approach (LSTM-GWO-FS) achieved 93% and 79%, compared to 82% and 77%. On the NSL-KDD dataset, GRU-GWO-FS improved accuracy to 94% and 92%, and LSTM-GWO-FS to 94% and 92% for anomaly and signature-based detection, respectively.Article Highlights: Novel and Efficient Approaches: Introduced two IDS models using Grey Wolf Optimization (GWO) with GRU and LSTM, significantly boosting detection accuracy and reducing computational overhead. Accuracy Gains: For the UNSW-NB15 dataset, GRU-GWO-FS achieved 90% anomaly detection and 79% signature detection, while LSTM-GWO-FS reached 93% and 79%. For the NSL-KDD dataset, both GRU-GWO-FS and LSTM-GWOFS attained 94% anomaly detection and 92% signature detection. [ABSTRACT FROM AUTHOR]

The classification and analysis of malicious software (malware) has reached a huge development in the systems associated with the internet. The malware exploits the system information and takes off the important information of the user without any intimation. Moreover, the malware furtively directs that information to the servers which are organized by the attackers. In recent years, many researchers and scientists discovered anti-malware products to identify known malware. But these methods are not robust to detect obfuscated and packed malware. To overcome these problems, the hybridized grasshopper optimization and cuckoo search (GOA-CSA) algorithm is proposed. The effective features are selected by the GOA-CSA algorithm which eases the process of classifying the malware. This research also utilized long short-term memory (LSTM)-softsign classifier to classify the malware. The malware samples are collected from the VXHeavens dataset which consists of malware samples from various software. The proposed model performance is estimated by using the performance metrics like accuracy, sensitivity, recall, and F1-score. The model attained better accuracy of 98.95% when the model is compared with other existing models. [ABSTRACT FROM AUTHOR]

Abstract Accurate carbon price forecasts are crucial for policymakers and enterprises to understand the dynamics of carbon price fluctuations, enabling them to formulate informed policies and investment strategies. However, due to the non-linear and non-stationary nature of carbon price, traditional models often struggle to achieve high prediction accuracy. To address this challenge, this study proposes a novel integrated prediction framework designed to enhance forecast accuracy. First, the carbon price series is decomposed into a series of smoother subsequences using fast iterative filtering (FIF). Subsequently, an integrated prediction model, AM-TCN-LSTM, is constructed, incorporating the attention mechanism (AM), temporal convolutional networks (TCN), and long short-term memory (LSTM) neural networks. The attention mechanism adaptively captures complex features from multiple factors, while the TCN-LSTM efficiently extracts temporal features from the sequences. Finally, the results from each subsequence are aggregated to generate the final prediction. Five carbon markets in china: Guangdong, Hubei, Shenzhen, Beijing, and Shanghai were selected to verify the validity of the proposed model. Various comparative models and evaluation metrics were employed to assess performance. The results demonstrate that: (1) the TCN-LSTM model achieves higher prediction accuracy compared to single models. (2) FIF is a more effective decomposition method with superior performance compared to EMD-based methods. (3) The proposed model exhibits the highest predictive capability, with MAE values of 0.0964, 0.1403, 1.9476, 2.0848, and 0.5029 for the five carbon markets, significantly outperforming comparison models. (4) The attention mechanism effectively captures the influence of multiple factors on carbon price, particularly within the short-term components.

Abstract In the rapidly evolving landscape of network communication systems, the need for robust security measures has become paramount due to increased vulnerability to cyber threats. Traditional Intrusion Detection Systems (IDSs) face challenges in efficiently handling redundant features, leading to increased computational complexity. This research addresses these challenges by proposing two optimized IDSs leveraging Grey Wolf Optimization (GWO) combined with deep learning (DL) models. The first system integrates Gated Recurrent Unit (GRU) with GWO (GRU-GWO), while the second utilizes Long Short-Term Memory (LSTM) with GWO (LSTM-GWO). These systems aim to enhance feature selection, reducing dimensionality and improving detection accuracy. The NSL-KDD and UNSW-NB15 datasets, representative of contemporary network environments, were employed to evaluate the proposed systems. Experimental results demonstrate significant improvements in intrusion detection accuracy and computational efficiency, underscoring the efficacy of the DL-GWO approach in enhancing network security. The first approach (GRU-GWO-FS) increased accuracy to 90% and 79% for anomaly and signature-based detection on the UNSW-NB15 dataset, compared to 80% and 77% with all features. The second approach (LSTM-GWO-FS) achieved 93% and 79%, compared to 82% and 77%. On the NSL-KDD dataset, GRU-GWO-FS improved accuracy to 94% and 92%, and LSTM-GWO-FS to 94% and 92% for anomaly and signature-based detection, respectively.

Accurate carbon price forecasts are crucial for policymakers and enterprises to understand the dynamics of carbon price fluctuations, enabling them to formulate informed policies and investment strategies. However, due to the non-linear and non-stationary nature of carbon price, traditional models often struggle to achieve high prediction accuracy. To address this challenge, this study proposes a novel integrated prediction framework designed to enhance forecast accuracy. First, the carbon price series is decomposed into a series of smoother subsequences using fast iterative filtering (FIF). Subsequently, an integrated prediction model, AM-TCN-LSTM, is constructed, incorporating the attention mechanism (AM), temporal convolutional networks (TCN), and long short-term memory (LSTM) neural networks. The attention mechanism adaptively captures complex features from multiple factors, while the TCN-LSTM efficiently extracts temporal features from the sequences. Finally, the results from each subsequence are aggregated to generate the final prediction. Five carbon markets in china: Guangdong, Hubei, Shenzhen, Beijing, and Shanghai were selected to verify the validity of the proposed model. Various comparative models and evaluation metrics were employed to assess performance. The results demonstrate that: (1) the TCN-LSTM model achieves higher prediction accuracy compared to single models. (2) FIF is a more effective decomposition method with superior performance compared to EMD-based methods. (3) The proposed model exhibits the highest predictive capability, with MAE values of 0.0964, 0.1403, 1.9476, 2.0848, and 0.5029 for the five carbon markets, significantly outperforming comparison models. (4) The attention mechanism effectively captures the influence of multiple factors on carbon price, particularly within the short-term components. [ABSTRACT FROM AUTHOR]

Abstract Electroencephalography (EEG) serves as a crucial neurodiagnostic tool by recording the electrical brain activity via attached electrodes on the patient’s head. While artificial intelligence (AI) exhibited considerable promise in medical diagnostics, its potential in the realm of neurodiagnostics remains underexplored. This research addresses this gap by proposing an innovative approach employing time-series classification of EEG data, leveraging long-short-term memory (LSTM) neural networks for the identification of abnormal brain activity, particularly seizures. To enhance the performance of the proposed model, metaheuristic algorithms were employed for optimizing hyperparameter collection. Additionally, a tailored modification of the variable neighborhood search (VNS) is introduced, specifically tailored for this neurodiagnostic application. The effectiveness of this methodology is evaluated using a carefully curated dataset comprising real-world EEG recordings from both healthy individuals and those affected by epilepsy. This software-based approach demonstrates noteworthy results, showcasing its efficacy in anomaly and seizure detection, even when working with relatively modest sample sizes. This research contributes to the field by illuminating the potential of AI in neurodiagnostics, presenting a methodology that enhances accuracy in identifying abnormal brain activities, with implications for improved patient care and diagnostic precision.

As digital crime continues to rise, the preservation of digital evidence has become a critical phase in digital forensic investigations. This phase focuses on securing and maintaining the integrity of evidence for legal proceedings. Existing solutions for evidence preservation, such as centralized storage systems and cloud frameworks, present challenges related to security and collaboration. In this paper, we propose a novel framework that addresses these challenges in the preservation phase of forensics. Our framework employs a combination of advanced technologies, including the following: (1) Segmenting evidence into smaller components for improved security and manageability, (2) Utilizing steganography for covert evidence preservation, and (3) Implementing blockchain to ensure the integrity and immutability of evidence. Additionally, we incorporate Long Short-Term Memory (LSTM) networks to enhance steganography in the evidence preservation process. This approach aims to provide a secure, scalable, and reliable solution for preserving digital evidence, contributing to the effectiveness of digital forensic investigations. An experiment using linguistic steganography showed that the LSTM autoencoder effectively generates coherent text from bit streams, with low perplexity and high accuracy. Our solution outperforms existing methods across multiple datasets, providing a secure and scalable approach for digital evidence preservation. [ABSTRACT FROM AUTHOR]

Human cognitive functions such as perception, attention, learning, memory, reasoning, and problem-solving are all significantly influenced by emotion. Emotion has a particularly potent impact on attention, modifying its selectivity in particular and influencing behavior and action motivation. Artificial Emotional Intelligence (AEI) technologies enable computers to understand a user's emotional state and respond appropriately. These systems enable a realistic dialogue between people and machines. The current generation of adaptive user interference technologies is built on techniques from data analytics and machine learning (ML), namely deep learning (DL) artificial neural networks (ANN) from multimodal data, such as videos of facial expressions, stance, and gesture, voice, and bio-physiological data (such as eye movement, ECG, respiration, EEG, FMRT, EMG, eye tracking). In this study, we reviewed existing literature based on ML and data analytics techniques being used to detect emotions in speech. The efficacy of data analytics and ML techniques in this unique area of multimodal data processing and extracting emotions from speech. This study analyzes how emotional chatbots, facial expressions, images, and social media texts can be effective in detecting emotions. PRISMA methodology is used to review the existing survey. Support Vector Machines (SVM), Naïve Bayes (NB), Random Forests (RF), Recurrent Neural Networks (RNN), Logistic Regression (LR), etc., are commonly used ML techniques for emotion extraction purposes. This study provides a new taxonomy about the application of ML in SER. The result shows that Long-Short Term Memory (LSTM) and Convolutional Neural Networks (CNN) are found to be the most useful methodology for this purpose. [ABSTRACT FROM AUTHOR]

The food and beverage industry significantly impacts the global economy, subject to various influential factors. This study aims to develop an AI-powered model to enhance the understanding of regional food and beverage sales dynamics with a primary goal of globalizing food items based on ingredient consumption metrics. Methodologically, this research employs Long-Short Term Memory (LSTM) architecture RNN to create a framework to predict food item performance using historical time series data. The model's hyperparameters are optimized using genetic algorithm (GA), resulting in higher accuracy and a more flexible model suitable for growing and real-time data. Data preprocessing involves comprehensive analysis, cleansing, and feature engineering, including the use of gradient boosting models with K-fold cross-validation for revenue prediction. Historical sales data from 1995 to 2014, sourced from Kaggle open-source database, are prepared to capture temporal dependencies using sliding window techniques, making it suitable for LSTM model input. Evaluation metrics reveal the hybrid LSTM-GA model's efficacy, outperforming baseline LSTM with an MSE reduction from 0.045 to 0.029. Ultimately, this research underscores the development of a model that harnesses historical sales data and sophisticated machine learning techniques to forecast food item sales growth, empowering informed investment decisions and strategic expansions in the global food market. [ABSTRACT FROM AUTHOR]

Background and objective: Epilepsy, which is associated with neuronal damage and functional decline, typically presents patients with numerous challenges in their daily lives. An early diagnosis plays a crucial role in managing the condition and alleviating the patients' suffering. Electroencephalogram (EEG)- based approaches are commonly employed for diagnosing epilepsy due to their effectiveness and non-invasiveness. In this study, a classification method is proposed that use fast Fourier Transform (FFT) extraction in conjunction with convolutional neural networks (CNN) and long short-term memory (LSTM) models. Methods: Most methods use traditional frameworks to classify epilepsy, we propose a new approach to this problem by extracting features from the source data and then feeding them into a network for training and recognition. It preprocesses the source data into training and validation data and then uses CNN and LSTM to classify the style of the data. Results: Upon analyzing a public test dataset, the top-performing features in the fully CNN nested LSTM model for epilepsy classification are FFT features among three types of features. Notably, all conducted experiments yielded high accuracy rates, with values exceeding 96% for accuracy, 93% for sensitivity, and 96% for specificity. These results are further benchmarked against current methodologies, showcasing consistent and robust performance across all trials. Our approach consistently achieves an accuracy rate surpassing 97.00%, with values ranging from 97.95 to 99.83% in individual experiments. Particularly noteworthy is the superior accuracy of our method in the AB versus (vs.) CDE comparison, registering at 99.06%. Conclusion: Our method exhibits precise classification abilities distinguishing between epileptic and non-epileptic individuals, irrespective of whether the participant's eyes are closed or open. Furthermore, our technique shows remarkable performance in effectively categorizing epilepsy type, distinguishing between epileptic ictal and interictal states versus non-epileptic conditions. An inherent advantage of our automated classification approach is its capability to disregard EEG data acquired during states of eye closure or eye-opening. Such innovation holds promise for real-world applications, potentially aiding medical professionals in diagnosing epilepsy more efficiently. [ABSTRACT FROM AUTHOR]

Knee joint contact forces are commonly estimated via surrogate measures (i.e., external knee adduction moments or musculoskeletal modeling). Despite its capabilities, modeling is not optimal for clinicians or persons with limited experience. The purpose of this study was to design a novel prediction method for knee joint contact forces that is simplistic in terms of required inputs. This study included marker trajectories and instrumented knee forces during normal walking from the “Grand Challenge” (n = 6) and “CAMS” (n = 2) datasets. Inverse kinematics were used to derive stance phase hip (sagittal, frontal, transverse), knee (sagittal, frontal), ankle (sagittal), and trunk (frontal) kinematics. A long-short term memory network (LSTM) was created using matlab to predict medial and lateral knee force waveforms using combinations of the kinematics. The Grand Challenge and CAMS datasets trained and tested the network, respectively. Musculoskeletal modeling forces were derived using static optimization and joint reaction tools in OpenSim. Waveform accuracy was determined as the proportion of variance and root-mean-square error between network predictions and in vivo data. The LSTM network was highly accurate for medial forces (R² = 0.77, RMSE = 0.27 BW) and required only frontal hip and knee and sagittal hip and ankle kinematics. Modeled medial force predictions were excellent (R² = 0.77, RMSE = 0.33 BW). Lateral force predictions were poor for both methods (LSTM R² = 0.18, RMSE = 0.08 BW; modeling R² = 0.21, RMSE = 0.54 BW). The designed LSTM network outperformed most reports of musculoskeletal modeling, including those reached in this study, revealing knee joint forces can accurately be predicted by using only kinematic input variables. [ABSTRACT FROM AUTHOR]

Deep learning methods have been widely used in applications of yield estimation using remote sensing data. However, they still face some challenges in terms of accuracy due to their inability to fully utilize the spatiotemporal information of remote sensing data. In response to this problem, a winter wheat yield estimation method based on ensemble deep learning and Gaussian process (GP) was proposed. First, the long-short term memory (LSTM) network and convolutional neural network (CNN) were constructed to explore deep spatiotemporal features from Sentinel-1A time-series images. Then, a GP component was applied for fusion of intraimage deep spatiotemporal features and inter-sample spatial consistency features. Finally, the yield estimation results were obtained. The experimental results showed that the proposed method had higher accuracy than those compared models, with a coefficient of determination (${R^2}$ R 2 ) of 0.698, a root mean square error (RMSE) of 477.045 kg/ha and a mean absolute error (MAE) of 404.377 kg/ha, demonstrating the application potential of the proposed method in crop yield estimation applications. [ABSTRACT FROM AUTHOR]

In cloud/fog-based environments, resource management is an important and challenging process. The deadline-based workflow scheduling mechanism is a common practice in such systems to overcome the complexities of resource management. However, many proposed approaches suffer from resource overloading/underloading, ignoring volunteer and volatile resources, and acting reactively. This paper presents a load-balancing method for IoT/Fog/Cloud environments integrated with local schedulers based on predicting workload and the presence of volatile mobile nodes (as dynamic resources). The proposed approach, firstly, turns the environment into a grid of equal-sized cells to reduce the system's complexity. Then, the overall status of intra-cell resources (overloaded, underloaded, or normal) is estimated. This estimation is done according to the workload prediction and available dynamic resources. Finally, an exhaustive search is applied to dispatch extra workflows from an overloaded cell to an underloaded one in such a way as to avoid missing workflow deadlines and improve system performance. The proposed method is intended to be scalable and decentralized by nature, allowing it to be used in large-scale settings such as smart cities. Extensive software simulation is used to evaluate and compare the proposed method to with two recently published works. The simulation results show that the proposed method outperforms others regarding job completion rate, workload variances, and time-related parameters. [ABSTRACT FROM AUTHOR]

Copyright of Cihan University-Erbil Scientific Journal is the property of Cihan University-Erbil and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Language is a fundamental medium for human communication, encompassing spoken and written forms, each governed by grammatical rules. Sindhi, one of the oldest languages, is characterized by its rich morphology and grammatical structure. Part-of-speech (POS) tagging, a crucial process in natural language processing, involves assigning grammatical tags to words. This research presents a novel approach to POS tagging for Sindhi text using deep learning techniques. We developed a POS tagger employing Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) models, with LSTM demonstrating superior effectiveness. This study represents the first application of these deep learning methods for POS tagging in Sindhi. Utilizing fastText, we trained 79,959 Sindhi word vectors, derived from a corpus compiled from diverse sources including Sindhi books, stories, and poetry. The corpus comprises 1,459 sentences and 10,584 unique words, split into 80% for training and 20% for validation. Our results indicate that the LSTM model achieved an accuracy of 85.80%, outperforming the GRU model, which achieved 80.77%, by a margin of 5%. This work's novelty lies in the application of deep learning techniques to enhance POS tagging accuracy in the Sindhi language corpus. [ABSTRACT FROM AUTHOR]

The prediction of groundwater level is a viable strategy for attaining sustainable water resource management. Recently, machine learning techniques have gained popularity as an alternative to numerical and statistical time series models grounded in physical principles. These methods excel at spotting complex trends and non-linear relationships. This study applies and compares seven machine learning techniques to predict the 20-year monthly groundwater level over the Mashhad plain aquifer in Iran. A novel idea based on the Thiessen polygon is proposed to provide a compressive dataset for presenting groundwater level of whole aquifer where the machine learning models can be trained and tested. This is because there are several piezometric wells, rainfall gauges, and hydrometric stations in the entire aquifer. Extensive simulations and modelling are conducted to select the appropriate input combinations for each technique, to identify the best model, and to carry out sensitivity analysis using a novel criterion known as the Global Performance Index, which integrates several regression performance criteria. The results show that the well-known Long Short-Term Memory performs significantly better than its competitors. Its superiority is about 13% over the next technique. To have a more accurate Long Short-Term Memory model, the sensitivity analysis was performed to reach the optimal parameters are as 120, 0.17, and 100 for the number of neurons, dropout rate, and batch size, respectively. Furthermore, an uncertainty analysis using Moving Block Bootstrap is performed to ensure that all uncertain effects are eliminated. [ABSTRACT FROM AUTHOR]

Abstract As the number of identified proteins has expanded, the accurate identification of proteins has become a significant challenge in the field of biology. Various computational methods, such as Support Vector Machine (SVM), K-nearest neighbors (KNN), and convolutional neural network (CNN), have been proposed to recognize deoxyribonucleic acid (DNA)-binding proteins solely based on amino acid sequences. However, these methods do not consider the contextual information within amino acid sequences, limiting their ability to adequately capture sequence features. In this study, we propose a novel approach to identify DNA-binding proteins by integrating a CNN with bidirectional long-short-term memory (LSTM) and gated recurrent unit (GRU) as (CNN-BiLG). The CNN-BiLG model can explore the potential contextual relationships of amino acid sequences and obtain more features than traditional models. Our experimental results demonstrate a validation set prediction accuracy of 94% for the proposed CNN-BiLG, surpassing the accuracy of machine learning models and deep learning models. Furthermore, our model is both effective and efficient, exhibiting commendable classification accuracy based on comparative analysis.

Over the last few years, air pollution has become a matter of great concern. Numerous machine learning and deep learning techniques have been applied to predict PM2.5 (Particulate Matter2.5). However, deterministic models perform forecasting based on the mean of probable outputs and cannot handle the uncertainties in real-life situations. With the aim of solving the low accuracy of PM2.5 concentration prediction during uncertainties, the present study proposed an innovative probabilistic model -Prob PM2.5 which predicts one day ahead PM2.5 concentration for time series data, which is multivariate in nature. First, a comprehensive correlation analysis between the meteorological features and PM2.5 concentration is done. Finally, the Conditional GAN framework is used to train the ProbPM2.5 with the help of adversarial training. The proposed framework that transformed a deterministic model into a probabilistic model provided improved performance. Comparative analysis with conventional models, such as LSTM (Long Short-Term Memory) and GRU (Gated Recurrent Unit) reveals that ProbPM2.5 outperforms during testing, showcasing resilience in the face of unforeseen events like COVID-19. Hence, the proposed method could perform improved characterization of time series characteristics of the air pollutant changes in order to obtain better accuracy of PM2.5 concentration prediction. [ABSTRACT FROM AUTHOR]

Owing to the dynamic nature of the financial industry, determining accurate stock market forecasts remains a significant challenge. Traditional forecasting methods often struggle to capture the intricate and volatile dynamics of stock price movements. Similarly numerous strategies for stock market prediction have been presented, precise prediction in this field still requires attention. Based on this insight, a novel sequential training model is proposed by adopting the optimal feature selection procedure. In order to determine stock price predictions, primarily financial Nifty data is obtained from the corresponding source. After acquiring financial data, the feature extraction phase is used to extract features from fundamental analysis, such as the Relative Strength Index, Rate of Change, Average True Range, and Exponential Moving Average. Additionally, statistical characteristics such as mean, standard deviation, variance, skewness, and kurtosis are derived from the stock market data. In order to select parameters, the fitness dependent randomised tunicate swarm optimization technique is utilized after the features have been retrieved. Feature selection improves the deep learning process and increases prediction capability by selecting the most important variables and eliminating irrelevant features. A novel sequential training technique is introduced aimed at forecasting stock market trends by leveraging the chosen features. The suggested approach undergoes comprehensive testing, evaluating its predictive capability using accuracy, precision, and recall metrics, implemented towards enhancing future stock price forecasts. [ABSTRACT FROM AUTHOR]

The detection of anomalous behavior in video surveillance is a focus of present research which has huge value and extensive application probabilities. Due to the difficulty of human movement and the feasibility of environments, anomalous behavior detection has some challenges. This paper proposed a deep learning-based approach for detecting the weapon and anomalous behavior. The Residual Attention-based Long-Short Term Memory (LSTM) with Self Gated Rectified Linear Unit (SGReLU) is proposed to enhance the detection accuracy. The median filter is used for preprocessing which removes noise from the UCF-Crime dataset and feeds into Histogram of Oriented Gradients (HOG) for feature extraction. Then, the Reverse Learning Chimp Optimization Algorithm (RL-COA) is utilized for hyperparameter optimization which attains the individual's reverse solution and then preserves the individual with high fitness values. At last, the Residual Attention-based LSTM with SGReLU is utilized for the classification process. This model overcomes neuron dead issues by allowing negative values for some neurons and minimizing the probability of inactive neurons. The proposed model attained better results on UCF-Crime dataset through the metrics like accuracy, precision, recall, f1-score and AUC values of about 98.84%, 98.62%, 98.47%, 98.35% and 98.21% correspondingly that ensures accurate detection when compared to existing techniques such as ResNet18 with Simple Recurrent Unit (SRU), Residual attention-based LSTM and Convolutional LSTM. [ABSTRACT FROM AUTHOR]

In the quest for sustainable urban development, precise quantification of urban green space is paramount. This research delineates the implementation of a Cosine Adaptive Particle Swarm Optimization Long Short-Term Memory (CAPSO-LSTM) model, utilizing a comprehensive dataset from Beijing (1998–2021) to train and test the model. The CAPSO-LSTM model, which integrates a cosine adaptive mechanism into particle swarm optimization, advances the optimization of long short-term memory (LSTM) network hyperparameters. Comparative analyses are conducted against conventional LSTM and Partical Swarm Optimization (PSO)-LSTM frameworks, employing mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE) as evaluative benchmarks. The findings indicate that the CAPSO-LSTM model exhibits a substantial improvement in prediction accuracy over the LSTM model, manifesting as a 66.33% decrease in MAE, a 73.78% decrease in RMSE, and a 57.14% decrease in MAPE. Similarly, when compared to the PSO-LSTM model, the CAPSO-LSTM model demonstrates a 58.36% decrease in MAE, a 65.39% decrease in RMSE, and a 50% decrease in MAPE. These results underscore the efficacy of the CAPSO-LSTM model in enhancing urban green space area prediction, suggesting its significant potential for aiding urban planning and environmental policy formulation. [ABSTRACT FROM AUTHOR]

Psychological activities have various dimensions in which they correlate with their respective behavior generated by the human body. Understanding the relationship of psychological events with the help of external action units is one of the research subjects to explore various human behavior and their dependencies. Existing work applied various deep learning algorithms to outline the correlation between psychological activities with human emotions. The study of psychological analysis in the medical field is very time-consuming and costly. It requires constant monitoring of the patient for a period of time and various interrogation sessions to finalize the emotional severity of an individual. Few skilled specialists and the lack of medical knowledge of human emotions drive the need for computer vision approaches to emphasize emotion recognition, particularly in disorders. The proposed study specifically assesses the use of speech signals to identify psychological disorders in terms of stress characteristics and uses a deep learning model that incorporates feed-forward networks and long short-term memory (LSTM) to identify the degree of psychological disease. The study made use of a standard speech dataset that was gathered from a variety of patients using a standard questionnaire format. The survey was conducted throughout a few Indian states. Speech samples were taken from patients whose cortisol levels were higher than 10 %. To assess the relationship between speech and psychological activity, speech signals from each patient have been gathered. The spectrogram of the speech signal's Mel filter bank coefficients has been analyzed, and the characteristics that cause stress and those that don't have it have been further divided into categories. The suggested model classifies stress and non-stress features in 150 voice dataset subjects with an average accuracy of 98 %. The model is found to be robust for various applications such as preventing suicidal cases, improving decision-making in the diagnosis of depression patients, improves the overall mental healthcare system.

Abstract The simulation of magnetic bearings involves highly non-linear physics, with high dependency on the input variation. Moreover, such a simulation is time consuming and can’t run, within realistic computation time for control purposes, when using classical computation methods. On the other hand, classical model reduction techniques fail to achieve the required precision within the allowed computation window. To address this complexity, this work proposes a combination of physics-based computing methods, model reduction techniques and machine learning algorithms, to tackle the requirements. The physical model used to represent the magnetic bearing is the classical Cauer Ladder Network method, while the model reduction technique is applied on the error of the physical model’s solution. Later on, in the latent space a machine learning algorithm is used to predict the evolution of the correction in the latent space. The results show an improvement of the solution without scarifying the computation time. The solution is computed in almost real-time (few milliseconds), and compared to the finite element reference solution.

Classification of volatile organic compounds (VOCs) is of interest in many fields. Examples include but are not limited to medicine, detection of explosives, and food quality control. Measurements collected with so-called electronic noses can be used for classification and analysis of VOCs. One type of electronic noses that has seen considerable development in recent years is Differential Mobility Spectrometry (DMS). DMS yields measurements that are visualized as dispersion plots that contain traces, also known as alpha curves. Current methods used for analyzing DMS dispersion plots do not usually utilize the information stored in the continuity of these traces, which suggests that alternative approaches should be investigated. In this work, for the first time, dispersion plots were interpreted as a series of measurements evolving sequentially. Thus, it was hypothesized that time-series classification algorithms can be effective for classification and analysis of dispersion plots. An extensive dataset of 900 dispersion plots for five chemicals measured at five flow rates and two concentrations was collected. The data was used to analyze the classification performance of six algorithms. The highest classification accuracy of 88% was achieved by a Long-Short Term Memory neural network, which supports the hypothesis that interpreting DMS measurements as sequential data is beneficial and outperformed classification algorithms traditionally used for DMS-based VOC identification.

In recent years, there has been evidence of a growing interest on the part of universities to know in advance the academic performance of their students and allow them to establish timely strategies to avoid desertion and failure. One of the biggest challenges to predicting student performance is presented in the course “Programming Fundamentals” of Computer Science, Software Engineering, and Information Systems Engineering careers in Peruvian universities for high student dropout rates. The objective of this research was to explore the efficiency of Long-Short Term Memory Networks (LSTM) in the field of Educational Data Mining (EDM) to predict the academic performance of students during the seventh, eighth, twelfth, and sixteenth weeks of the academic semester, which allowed us to identify students at risk of failing the course. This research compares several predictive models, such as Deep Neural Network (DNN), Decision Tree (DT), Random Forest (RF), Logistic Regression (LR), Support Vector Classifier (SVM), and K-Nearest Neighbor (KNN). A major challenge machine learning algorithms face is a class imbalance in a dataset, resulting in over-fitting to the available data and, consequently, low accuracy. We use Generative Adversarial Networks (GAN) and Synthetic Minority Over-sampling Technique (SMOTE) to balance the data needed in our proposal. From the experimental results based on accuracy, precision, recall, and F1-Score, the superiority of our model is verified concerning a better classification, with 98.3% accuracy in week 8 using LSTM-GAN, followed by DNN-GAN with 98.1% accuracy.

Membrane desalination (MD) is an efficient process for desalinating saltwater, combining the uniqueness of both thermal and separation distillation configurations. In this context, the optimization strategies and sizing methodologies are developed from the balance of the system’s energy demand. Therefore, robust prediction modeling of the thermodynamic behavior and freshwater production is crucial for the optimal design of MD systems. This study presents a new advanced machine-learning model to obtain the permeate flux of a tubular direct contact membrane distillation unit. The model was established by optimizing a long-short-term memory (LSTM) model by an election-based optimization algorithm (EBOA). The model inputs were the temperatures of permeate and the feed flow, and the rate and salinity of the feed flow. The optimized model was compared with other optimized LSTM models by sine–cosine optimization algorithm (SCA), artificial ecosystem optimizer (AEO), and grey wolf optimization algorithm (GWO). All models were trained, tested, and evaluated using different accuracy measures. LSTM-EBOA outperformed other models in predicting the permeate flux based on different accuracy measures. LSTM-EBOA had the highest coefficient of determination of 0.998 and 0.988 and the lowest root mean square error of 1.272 and 4.180 for training and test, respectively. It can be recommended that this paper provide a useful pathway for sizing parameters selection and predicting the performance of MD systems that makes an optimally designed model for predicting the freshwater production rates without costly experiments.

Accurate prediction of the Remaining Useful Life (RUL) of components and systems is crucial for avoiding an unscheduled shutdown of production by planning maintenance interventions effectively in advance. For high‐reliability equipment, few complete‐run‐to‐failure trajectories may be available in practice. This constitutes a technical challenge for data‐driven techniques for estimating the RUL. This paper proposes a novel data‐driven approach for fault prognostics using the Long‐Short Term Memory (LSTM) model combined with the Multi‐Head Self‐Attention (MHSA) mechanism. The former is applied to the input signals, whereas the latter is used to extract features from the LSTM hidden states, benefiting from the information from all hidden states rather than utilizing that of the final hidden state only. The proposed approach is characterized by its capability to recognize long‐term dependencies while extracting features in both global and local contexts. This enables the approach to provide accurate RUL estimates in various stages of the equipment's life. The proposed approach is applied to an artificial case study simulated to mimic the realistic degradation behaviour of a heterogeneous fleet of aluminium electrolytic capacitors used in the automotive industry (under variable operating and environmental conditions). Results indicate that the proposed approach can provide accurate RUL estimates for high‐reliability equipment compared to four benchmark models from the literature. [ABSTRACT FROM AUTHOR]

Power system oscillation is an unavoidable threat to the stability of an interconnected modern power system. The reliable operation of a modern power system is widely related to the dampening of electromechanical low‐frequency oscillations (ELFOs). These ELFOs must be dampened appropriately to maintain the stability and reliability of the system. However, it is relatively difficult to resolve the problem of ELFOs completely with traditional power system stabilizers (PSSs). Consequently, research should be directed towards the development of efficient damping controllers, or PSSs, for power oscillation damping. Motivated from the aforementioned fact, this article presents the design of a proportional integral derivative power system stabilizer (PID‐PSS) via a long short‐term memory neural network (LSTM) based deep neural approach for damping power system oscillations in interconnected power systems. LSTM is used to train the parameters of PID‐PSS. To evaluate the performance of the proposed LSTM based PID‐PSS, diverse test cases under different operating conditions are examined. Further, the performance of the proposed LSTM based PID‐PSS is compared with traditional PSSs through time‐domain simulations. The test cases reveal the desired efficiency achieved by the proposed LSTM based PID‐PSS under diverse loading conditions. [ABSTRACT FROM AUTHOR]

In our paper we intend to present a methodology that we elaborated for clustering texts based on the word fre quency in the English translations of selected old Greek texts. We used the classification system of the ancient Library of Alex andria, devised by the prominent Greek scholar-poet, Callima chus in the 3rd century BC., as a basis for categorizing literary masterpieces. In our content analysis, we could determine a tri plet of a, b, c values for describing a power function that appro priately fits a curve determined by the word frequencies in the texts. In addition, we have discovered 16 special features of the different texts that correspond to various token categories inves tigated in each text, such as part of speech of the word in the con text, numerals, subordinate conjunction, symbols, etc. We have developed a cognitive model in which several hundred different subtexts were utilized for supervised learning with the aim of subtext class recognition. Concerning 200 subtexts, the triplet of a, b, c values, the classes of the subtexts, and their 16-dimen sional feature vectors were learnt for the Recurrent Neural Net work (RNN). It turned out that the Long-Short Term Memory RNN could efficiently predict which class a chosen subtext could be categorized into without considering the interpretation of the content. The influence of the non-zero error rate of new com munication services on the meaning of the transferred texts was also investigated. The impact of the noise on the classification accuracy was found to be linear, dependent on the character error rate. [ABSTRACT FROM AUTHOR]

TBM tunnelling attitude controlling is a significant issue for guaranteeing the tunnel fitting the expected tunnel axis, with directly influence the tunnel quality. The key to solve the problem is to establish the relationship between the tunnelling attitude and the controlling parameters and to predict the tunnelling attitude accordingly. For this, this paper introduced a TBM tunnelling attitude predicting method. In detail, using Long-Short Term Memory (LSTM), the initial tunnelling attitude and the controlling parameters of each later ring are taken as input, while the tunnelling attitude of each later rings are regarded as the output, and the relationship between the input and output is established. Meanwhile, for avoid the over-fitting and error accumulation risk of LSTM, the theoretical relationship between the input and output is also built based on the TBM mechanical movement principle, and it is also involved into the LSTM-based relationship as constraints. The proposed method is verified by the field data collected from the 6th Section of the Qingdao Metro Project, and the results reveal that the proposed LSTM-based method is accurate and acceptable. [ABSTRACT FROM AUTHOR]

The prediction of power demand for unmanned aerial vehicles (UAV) is an essential basis to ensure the rational distribution of the energy system and stable economic flight. In order to accurately predict the demand power of oil-electric hybrid UAV, a method based on variational mode decomposition (VMD) and Sparrow Search Algorithm (SSA) is proposed to optimize the hybrid prediction model composed of long-short term memory (LSTM) and Least Squares Support Vector Machine (LSSVM). Firstly, perform VMD decomposition on the raw demand power data and use the sample entropy method to classify the feature-distinct mode components into high-frequency and low-frequency categories. Then, each modality component was separately input into the mixed model for rolling prediction. The LSSVM model and LSTM model were used to process low-frequency and high-frequency components, respectively. Finally, the predicted values for each modal component are linearly combined to obtain the final predicted value for power demand. Compared with the current models, the prediction model constructed in this paper stands out for its superior ability to track the changing trends of power demand and achieve the highest level of prediction accuracy. [ABSTRACT FROM AUTHOR]

Flettner rotors are wind propulsion systems using the Magnus effect to generate thrust, thereby reduce fuel consumption and carbon emissions in the ships. However, rotor unbalance can cause excessive vibrations and energy loss, affecting the performance and stability of the system. There is a need to have a system onboard, which can predict the vibrations. The paper proposes a deep learning approach to predict the vibrations and unbalanced forces of a Flettner rotor based on the data of ECO Flettner rotor onboard the vessel MV Fehn pollux. The paper develops two methods to estimate the direction and magnitude of the unbalanced forces using the reading values of the strain gauges. The work also compares two recurrent neural network models, namely Long-short term memory and Gated Recurrent Unit, for vibration prediction and evaluates their performance using Mean Absolute Error and Root Mean Squared Error metrics. The results show that Longshort term memory model outperforms Gated Recurrent Unit model in prediction accuracy and can be implemented on the system onboard to monitor and prevent rotor unbalance. The paper also suggests some possible solutions for automatic self-balancing of the rotor and identifies some areas for future work. [ABSTRACT FROM AUTHOR]

Membrane desalination (MD) is an efficient process for desalinating saltwater, combining the uniqueness of both thermal and separation distillation configurations. In this context, the optimization strategies and sizing methodologies are developed from the balance of the system's energy demand. Therefore, robust prediction modeling of the thermodynamic behavior and freshwater production is crucial for the optimal design of MD systems. This study presents a new advanced machine-learning model to obtain the permeate flux of a tubular direct contact membrane distillation unit. The model was established by optimizing a long-short-term memory (LSTM) model by an election-based optimization algorithm (EBOA). The model inputs were the temperatures of permeate and the feed flow, and the rate and salinity of the feed flow. The optimized model was compared with other optimized LSTM models by sine–cosine optimization algorithm (SCA), artificial ecosystem optimizer (AEO), and grey wolf optimization algorithm (GWO). All models were trained, tested, and evaluated using different accuracy measures. LSTM-EBOA outperformed other models in predicting the permeate flux based on different accuracy measures. LSTM-EBOA had the highest coefficient of determination of 0.998 and 0.988 and the lowest root mean square error of 1.272 and 4.180 for training and test, respectively. It can be recommended that this paper provide a useful pathway for sizing parameters selection and predicting the performance of MD systems that makes an optimally designed model for predicting the freshwater production rates without costly experiments. [ABSTRACT FROM AUTHOR]

In the quest for sustainable urban development, precise quantification of urban green space is paramount. This research delineates the implementation of a Cosine Adaptive Particle Swarm Optimization Long Short-Term Memory (CAPSO-LSTM) model, utilizing a comprehensive dataset from Beijing (1998–2021) to train and test the model. The CAPSO-LSTM model, which integrates a cosine adaptive mechanism into particle swarm optimization, advances the optimization of long short-term memory (LSTM) network hyperparameters. Comparative analyses are conducted against conventional LSTM and Partical Swarm Optimization (PSO)-LSTM frameworks, employing mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE) as evaluative benchmarks. The findings indicate that the CAPSO-LSTM model exhibits a substantial improvement in prediction accuracy over the LSTM model, manifesting as a 66.33% decrease in MAE, a 73.78% decrease in RMSE, and a 57.14% decrease in MAPE. Similarly, when compared to the PSO-LSTM model, the CAPSO-LSTM model demonstrates a 58.36% decrease in MAE, a 65.39% decrease in RMSE, and a 50% decrease in MAPE. These results underscore the efficacy of the CAPSO-LSTM model in enhancing urban green space area prediction, suggesting its significant potential for aiding urban planning and environmental policy formulation.

Electric consumption prediction methods are investigated for many reasons, such as decision-making related to energy efficiency as well as for anticipating demand and the dynamics of the energy market. The objective of the present work is to compare two Deep Learning models, namely the Long Short-Term Memory (LSTM) model, and the Bi-directional LSTM (BLSTM) for univariate electric consumption Time Series (TS) short-term forecast model. The Data Sets (DSs) were selected for their different contexts and scales, with the goal of assessing the robustness of the models. Four DSs were used, related to the power consumption of: (a) a household in France; (b) a university building in Santarém, Brazil; (c) the Tétouan city zones, in Morocco; and (d) the aggregated electric demand of Singapore. The metrics RMSE, MAE, MAPE and R2 were calculated in a TS cross-validation scheme. Friedman’s test was applied to normalized RMSE (NRMSE) results, showing that BLSTM outperforms LSTM with statistically significant difference (p = 0.0455), corroborating the fact that bidirectional weight updating significantly improves the LSTM performance with respect to different scales of electric power consumption. The present work provides statistical evidence supporting the conclusion that BLSTM outperforms LSTM models according to the tests performed, based on a complete methodology for TS prediction, and also establishes a baseline for future investigation of electric consumption TS prediction.

Detecting and monitoring faults in solar photovoltaic (PV) systems is crucial to ensure optimal efficiency and prevent safety and fire hazards. However, the conventional operation and maintenance (O&M) of solar PV systems do not utilize machine learning for fault detection and classification. This poses challenges for plant operators, especially those managing large-scale solar (LSS) PV plants, who typically rely on manual approaches to screen large amounts of electrical data and inspect numerous string panels. Consequently, the cost of O&M is high. This study aims to use advanced machine-learning techniques to detect anomalies in a Large-Scale Photovoltaic (LSSPV) plant. The study collects data from the plant which located in the central of Peninsular Malaysia and employs K-Means for clustering and Long-Short Term Memory (LSTM) for anomaly detection in the predicted electrical current of string modules. The model is developed using Jupyter Notebook from the Python Package Index. To validate the accuracy of the proposed model, the study compares LSTM with Artificial Neural Network (ANN) using relative error as the evaluation metric. The results indicate that the LSTM method identifies anomalies in the predicted output current of the string modules more accurately and with lower relative error than the conventional ANN technique. The proposed model could help plant operators perform predictive maintenance of LSSPV plants at a minimal cost and time.

The accuracy of long-term runoff models can be increased through the input of local weather variables and global climate indices. However, existing methods do not effectively extract important information from complex input factors across various temporal and spatial dimensions, thereby contributing to inaccurate predictions of long-term runoff. In this study, local–global–temporal attention mechanisms (LGTA) were proposed for capturing crucial information on global climate indices on monthly, annual, and interannual time scales. The graph attention network (GAT) was employed to extract geographical topological information of meteorological stations, based on remotely sensed elevation data. A long-term runoff prediction model was established based on long-short-term memory (LSTM) integrated with GAT and LGTA, referred to as GAT–LGTA–LSTM. The proposed model was compared to five comparative models (LGTA–LSTM, GAT–GTA–LSTM, GTA–LSTM, GAT–GA–LSTM, GA–LSTM). The models were applied to forecast the long-term runoff at Luning and Pingshan stations in China. The results indicated that the GAT–LGTA–LSTM model demonstrated the best forecasting performance among the comparative models. The Nash–Sutcliffe Efficiency (NSE) of GAT–LGTA–LSTM at the Luning and Pingshan stations reached 0.87 and 0.89, respectively. Compared to the GA–LSTM benchmark model, the GAT–LGTA–LSTM model demonstrated an average increase in NSE of 0.07, an average increase in Kling–Gupta Efficiency (KGE) of 0.08, and an average reduction in mean absolute percent error (MAPE) of 0.12. The excellent performance of the proposed model is attributed to the following: (1) local attention mechanism assigns a higher weight to key global climate indices at a monthly scale, enhancing the ability of global and temporal attention mechanisms to capture the critical information at annual and interannual scales and (2) the global attention mechanism integrated with GAT effectively extracts crucial temporal and spatial information from precipitation and remotely-sensed elevation data. Furthermore, attention visualization reveals that various global climate indices contribute differently to runoff predictions across distinct months. The global climate indices corresponding to specific seasons or months should be selected to forecast the respective monthly runoff.

The deep learning diagnosis of aircraft engine-bearing faults enables cost-effective predictive maintenance while playing an important role in increasing the safety, reliability, and efficiency of aircraft operations. Because of highly dynamic and harsh operating conditions of this system, such modeling is challenging due to data complexity and drift, making it difficult to reveal failure patterns. As a result, the objective of this study is dual. To begin, a highly structured data preprocessing strategy ranging from extraction, denoising, outlier removal, scaling, and balancing is provided to solve data complexity that resides specifically in outliers, noise, and data imbalance problems. Gap statistics under k-means clustering are used to evaluate preprocessing results, providing a quantitative estimate of the ideal number of clusters and thereby enhancing data representations. This is the first time, to the best of authors' knowledge, that such a criterion has been employed for an important step in a preliminary ground truth validation in supervised learning. Furthermore, to tackle data drift issues, long-short term memory (LSTM) adaptive learning features are used and subjected to a learning parameter improvement method utilizing recursive weights initialization (RWI) across several rounds. The strength of such methodology can be seen by application to realistic, extremely new, complex, and dynamic data collected from a real test-bench. Cross validation of a single LSTM layer model with only 10 neurons shows its ability to enhance classification performance by 7.7508% over state-of-the-art results, obtaining a classification accuracy of 92.03 ± 0.0849%, which is an exceptional performance in such a benchmark. [ABSTRACT FROM AUTHOR]

Today, integrating wind energy forecasting is an important area of research due to the erratic nature of wind. To achieve this goal, we propose a new model of wind speed prediction based on graph attention networks (GAT), we added a new attention mechanism and a learnable adjacency matrix to the GAT structure to obtain attention scores for each weather variable. The results of the GAT-based model are merged with the bi-directional deep learning long and short-term memory (BiLSTM) layer to take advantage of the geographic and temporal properties of historical weather data. The experiments and analyzes are carried out using precise meteorological data collected from wind farms in the Moroccan city of Tetouan. We show that the proposed model can learn complex input-output correlations of meteorological data more efficiently than previous wind speed prediction algorithms. Due to the resulting attention weights, the model also provides more information about the main weather factors for the evaluated forecast work. [ABSTRACT FROM AUTHOR]

As a core component of small unmanned aerial vehicles (UAVs), GPS is playing a critical role in providing localization for UAV navigation. UAVs are an important factor in the large-scale deployment of the Internet of Things (IoT) and cyber–physical systems (CPS). However, GPS is vulnerable to spoofing attacks that can mislead a UAV to fly into a sensitive area and threaten public safety and private security. The conventional spoofing detection methods need too much overhead, which stops efficient detection from working in a computation-constrained UAV and provides an efficient response to attacks. In this paper, we propose a novel approach to obtain a lightweight detection model in the UAV system so that GPS spoofing attacks can be detected from a long distance. With long-short term memory (LSTM), we propose a lightweight detection model on the ground control stations, and then we distill it into a compact size that is able to run in the control system of the UAV with knowledge distillation. The experimental results show that our lightweight detection algorithm runs in UAV systems reliably and can achieve good performance in GPS spoofing detection. [ABSTRACT FROM AUTHOR]

Intelligent transportation systems (ITS) are gaining attraction in large cities for better traffic management. Traffic forecasting is an important part of ITS, but a difficult one due to the intricate spatiotemporal relationships of traffic between different locations. Despite the fact that remote or far sensors may have temporal and spatial similarities with the predicting sensor, existing traffic forecasting research focuses primarily on modeling correlations between neighboring sensors while disregarding correlations between remote sensors. Furthermore, existing methods for capturing spatial dependencies, such as graph convolutional networks (GCNs), are unable to capture the dynamic spatial dependence in traffic systems. Self-attention-based techniques for modeling dynamic correlations of all sensors currently in use overlook the hierarchical features of roads and have quadratic computational complexity. Our paper presents a new Dynamic Graph Convolution LSTM Network (DyGCN-LSTM) to address the aforementioned limitations. The novelty of DyGCN-LSTM is that it can model the underlying non-linear spatial and temporal correlations of remotely located sensors at the same time. Experimental investigations conducted using four real-world traffic data sets show that the suggested approach is superior to state-of-the-art benchmarks by 25 % in terms of RMSE. [ABSTRACT FROM AUTHOR]

This research aims to predict new COVID-19 cases in Bandung, Indonesia. The system implemented two types of deep learning methods to predict this. They were the recurrent neural networks (RNN) and long-short-term memory (LSTM) algorithms. The data used in this study were the numbers of confirmed COVID-19 cases in Bandung from March 2020 to December 2020. Pre-processing of the data was carried out, namely data splitting and scaling, to get optimal results. During model training, the hyperparameter tuning stage was carried out on the sequence length and the number of layers. The results showed that RNN gave a better performance. The test used the RMSE, MAE, and R2 evaluation methods, with the best numbers being 0.66975075, 0.47075, 0.29616625, and 0.7644 on the test data. [ABSTRACT FROM AUTHOR]