Short-term Memory Network Research Articles

Developing an Innovative Seq2Seq Model to Predict the Remaining Useful Life of Low-Charged Battery Performance Using High-Speed Degradation Data

This study introduces a novel Sequence-to-Sequence (Seq2Seq) deep learning model for predicting lithium-ion batteries’ remaining useful life. We address the challenge of extrapolating battery performance from high-rate to low-rate charging conditions, a significant limitation in previous studies. Experiments were also conducted on commercial cells using charge rates from 1C to 3C. Comparative analysis of fully connected neural networks, convolutional neural networks, and long short-term memory networks revealed their limitations in extrapolating to untrained conditions. Our Seq2Seq model overcomes these limitations, predicting charging profiles and discharge capacity for untrained, low-rate conditions using only high-rate charging data. The Seq2Seq model demonstrated superior performance with low error and high curve-fitting accuracy for 1C and 1.2C untrained data. Unlike traditional models, it predicts complete charging profiles (voltage, current, temperature) for subsequent cycles, offering a comprehensive view of battery degradation. This method significantly reduces battery life testing time while maintaining high prediction accuracy. The findings have important implications for lithium-ion battery development, potentially accelerating advancements in electric vehicle technology and energy storage.

Lower Limb Motion Recognition Based on sEMG and CNN-TL Fusion Model.

To enhance the classification accuracy of lower limb movements, a fusion recognition model integrating a surface electromyography (sEMG)-based convolutional neural network, transformer encoder, and long short-term memory network (CNN-Transformer-LSTM, CNN-TL) was proposed in this study. By combining these advanced techniques, significant improvements in movement classification were achieved. Firstly, sEMG data were collected from 20 subjects as they performed four distinct gait movements: walking upstairs, walking downstairs, walking on a level surface, and squatting. Subsequently, the gathered sEMG data underwent preprocessing, with features extracted from both the time domain and frequency domain. These features were then used as inputs for the machine learning recognition model. Finally, based on the preprocessed sEMG data, the CNN-TL lower limb action recognition model was constructed. The performance of CNN-TL was then compared with that of the CNN, LSTM, and SVM models. The results demonstrated that the accuracy of the CNN-TL model in lower limb action recognition was 3.76%, 5.92%, and 14.92% higher than that of the CNN-LSTM, CNN, and SVM models, respectively, thereby proving its superior classification performance. An effective scheme for improving lower limb motor function in rehabilitation and assistance devices was thus provided.

Deep learning-based ground motion inversion through recursive structural acceleration response using DRA-LSTM Net

Known ground motion time history is crucial for the precise numerical analysis of structural response and evaluation of its safety. This study introduces the Deep Recursive Attentive Long Short-Term Memory Network (DRA-LSTM Net), a framework designed for the inversion of ground motion (GM) from recorded structural response (SR) during an earthquake. By integrating an attention mechanism in network along with a strategic recursive approach for dataset management, the DRA-LSTM Net achieves high precision in GM inversion. The method incorporates a novel pre-processed matrix format to capture unique recursive behaviour patterns in GM data, enhancing the model’s training, validation efficiency, and prediction accuracy on unseen test datasets. Case studies on single-degree-of-freedom (SDOF) and multi-degree-of-freedom (MDOF) structures highlight the method’s potential for real-time and high-precision GM inversion. Additionally, transfer learning was implemented to fine-tune the pre-trained model with real-world sensor data and structural responses from different floors, demonstrating the model’s adaptability and ability to retain high prediction accuracy across diverse datasets. This approach is particularly useful for regions with limited seismic instrumentation but equipped with structural response sensors.

APPLICATION OF ARTIFICIAL INTELLIGENCE TECHNIQUES FOR ERROR PREVENTION AND QUALITY CONTROL IN MIG/MAG WELDING PROCESSES: A COMPREHENSIVE REVIEW

Welding techniques, particularly MIG/MAG (Metal Inert Gas/Metal Active Gas) processes, are essential for various industries, including automotive, construction, aerospace, and defense. Ensuring the quality and accuracy of welds is crucial for efficient production and product reliability. However, numerous issues can arise during MIG/MAG welding, often resulting in suboptimal quality. This study presents a comprehensive review of research efforts employing AI techniques to address errors and defects in MIG/MAG welding. The review systematically analyzes relevant studies published since 2018, focusing on three key aspects: datasets employed, methodologies and approaches adopted, and performance metrics reported. The findings reveal a significant adoption of both machine learning and deep learning techniques, with the choice of approach dependent on factors such as the nature of input data, welding process dynamics, and computational requirements. Deep learning models, particularly convolutional neural networks (CNNs) and long short-term memory (LSTM) networks, have demonstrated superior performance in tasks like image-based defect detection and time-series analysis for quality prediction. However, traditional machine learning algorithms have also been utilized, often in conjunction with dimensionality reduction or feature selection techniques. The review highlights the diverse range of performance metrics employed, including accuracy, precision, recall, F1-score, mean squared error (MSE), and root mean squared error (RMSE), with the selection contingent upon the specific task (classification or regression) and desired trade-off between different performance aspects. While many studies have reported promising results, with accuracy rates frequently exceeding 90%, challenges remain in real-world industrial settings, where factors such as noise, occlusions, and rapidly changing welding conditions can pose significant hurdles. This review serves as a comprehensive guide for researchers and practitioners working in the field of AI-assisted error prevention and quality control for MIG/MAG welding processes, highlighting current trends, methodologies, and future research directions.

Aplicación de redes neuronales artificiales a la modelación de lluvia-escorrentía en la cuenca del río Chancay Lambayeque

Between the months of December to April, regions of northern Peru, including Lambayeque, are affected by maximum extreme events, wreaking havoc on homes, flooding crop fields, collapsing hydraulic works, and the most irreparable loss of human lives. In this line, the objective of this research was to apply Artificial Neural Networks to rain-runoff modeling in a basin in northern Peru, namely, the Chancay Lambayeque river basin belonging to the Pacific slope. For this purpose, records of precipitation and flows of 30 years (hydrological normal) were collected from 12 hydrometeorological stations belonging to the basin and neighboring it. Thus, applying a model of Long and Short Term Memory Networks (LSTM) we proceeded to model the rain, seeking to follow the behavior of the flows observed in the Racarrumi hydrometric station, with 80 % of the information the model was trained and with 20 % it was validated. In short, it was obtained that in the modeling validation stage, the Nash coefficient was 0.93, corresponding to the qualifier "very good".

A day-ahead wind speed correction method: Enhancing wind speed forecasting accuracy using a strategy combining dynamic feature weighting with multi-source information and dynamic matching with improved similarity function

Forecasting error of day-ahead wind speed (WS) seriously affects wind power integration and power system security and stability. In this regard, this paper fully considers the spatiotemporal correlation of wind farms (WFs) in different geographical locations, and proposes a day-ahead WS combined correction method that integrates multi-source station dynamic information weighting. Different from the previous WS correction methods, this paper fully considers the dynamic correlation of WS between the WFs, introduces an improved weighted similarity function to screen and dynamically weight the information of WFs with dynamic correlation, and introduces the dynamic weighting feature into the WS correction process. A combined decomposition mechanism is proposed, which combines sequential variational mode decomposition (SVMD) and feature mode decomposition (FMD) models to extract the most relevant trend components and non-stationary components of forecasted and measured WS. A combined correction model is introduced, and a combined architecture of Non-stationary Transformer combined with bidirectional long short-term memory network (Ns-Transformer-BILSTM) is used to correct the stationary WS component. A dynamic matching mechanism of fluctuation components considering improved similarity is proposed for the correction of non-stationary components. The proposed method is applied to several regional WFs in China. The experimental results show that the average correction of NRMSE, NMAE and R can reach 2.4 % ∼ 3.7 %, 2.0 % ∼ 3.0 % and 3.3 % ∼ 9.7 %, respectively. The NRMSE and NMAE corresponding to the corrected WS of certain individual WFs can be reduced by 10 % and 9 %, respectively, and R can be increased by 33 %.

Hybrid deep learning model with VMD-BiLSTM-GRU networks for short-term traffic flow prediction

Accelerating urbanization and the rapid development of intelligent transportation systems have rendered short-term traffic flow prediction an important research field. Accurate prediction of traffic flow is beneficial for the optimization of traffic planning, improvement of road utilization, reduction of traffic congestion, and reduction in the incidence of traffic accidents. However, data pertaining to traffic flow are typically influenced by a multitude of factors, resulting in data that exhibit a considerable degree of nonlinearity and complexity. To address the issue of noise in raw traffic flow data, this study proposes a hybrid model that combines variational modal decomposition (VMD), a bidirectional long short-term memory network (BiLSTM), and a gated recurrent unit (GRU) for short-term traffic flow prediction. To validate the effectiveness of the model, an experimental validation was conducted based on traffic flow data from UK highways, and the performance of the model was compared with common benchmark models. The experimental results demonstrate that the proposed method yields superior prediction results in terms of mean absolute error, coefficient of determination, and root-mean-square error compared to existing prediction techniques, thereby substantiating its efficacy in short-term traffic flow prediction.

Riding Feeling Recognition Based on Multi-Head Self-Attention LSTM for Driverless Automobile

With the emergence of driverless technology, passenger ride comfort has become an issue of concern. In recent years, driving fatigue detection and braking sensation evaluation based on EEG signals have received more attention, and analyzing ride comfort using EEG signals is also a more intuitive method. However, it is still a challenge to find an effective method or model to evaluate passenger comfort. In this paper, we propose a long- and short-term memory network model based on a multiple self-attention mechanism for passenger comfort detection. By applying the multiple attention mechanism to the feature extraction process, more efficient classification results are obtained. The results show that the long- and short-term memory network using the multi-head self-attention mechanism is efficient in decision making along with higher classification accuracy. In conclusion, the classifier based on the multi-head attention mechanism proposed in this paper has excellent performance in EEG classification of different emotional states, and has a broad development prospect in brain-computer interaction.

Deep learning for prediction and classifying the dynamical behaviour of piecewise-smooth maps

This paper explores novel ways of predicting and classifying the dynamics of piecewise smooth maps using various deep learning models. Moreover, we have used machine learning models such as Decision Tree Classifier, Logistic Regression, K-Nearest Neighbor, Random Forest, and Support Vector Machine for predicting the border collision bifurcation in the 1D normal form map and the 1D tent map. The decision tree classifier best predicts the border collision bifurcation for the 1D normal form map, the random forest, and the 1D tent map. This study introduces a novel application of deep learning models to cobweb diagrams and phase portraits, which provides a new perspective for classifying regular and chaotic behaviour. Further, we classified the regular and chaotic behaviour of the 1D tent map and the 2D Lozi map using deep learning models like Convolutional Neural Network (CNN), ResNet50, and ConvLSTM via cobweb diagram and phase portraits, where CNN exhibits better performance than other models. We also classified the chaotic and hyperchaotic behaviour of the 3D piecewise smooth map using deep learning models such as the Feed Forward Neural Network (FNN), Long Short-Term Memory (LSTM), and Recurrent Neural Network (RNN). We have shown that LSTM performs best for classifying chaotic and hyperchaotic behaviour. Additionally, LSTM outperforms other models in accuracy and computational efficiency, making it highly effective for real-time analysis. Finally, deep learning models such as Long Short Term Memory (LSTM) and Recurrent Neural Network (RNN) are used for reconstructing the two-parameter bifurcation charts of 2D normal form map, in which LSTM is more precise than RNN in reconstructing the two-parametric charts.

Image-driven prediction system: Automatic extraction of aggregate gradation of pavement core samples integrating deep learning and interactive image processing framework

Aggregate gradation plays a crucial role in determining the asphalt pavement performance, necessitating assessment post-construction. Additionally, efficient gradation measurement is vital for the design and construction of reclaimed pavement. This study proposes a methodology for predicting gradation based on imagery. Initially, a scanning device is devised to capture unfolded side surface images of pavement core samples. Subsequently, an interactive software named PyImageJ is developed for image post-processing and acquiring aggregate feature information. A morphological model is then introduced to determine the mass ratios among various grading levels of coarse aggregates with nominal particle sizes greater than 2.36 mm. Finally, sequence prediction models are established to forecast the mass ratios among different grading levels of fine aggregates with nominal particle sizes smaller than 2.36 mm. Simple conversion of the mass ratios can yield the gradation. Experimental results indicate that the Long Short-Term Memory (LSTM) network offers more accurate predictions for the mass ratios of fine aggregates in stone mastic asphalt (SMA-13), with an R-squared value of 0.8985. Conversely, one-dimensional convolutional neural networks (1D CNN) are better suited for predicting the fine aggregate mass ratios in asphalt concrete (AC-20 and AC-25), with R-squared values of 0.9010 and 0.8824, respectively. Visualization results demonstrate that the predicted complete gradation closely aligns with the actual values (R-squared > 0.99), opening a new avenue for evaluating pavement quality.

Dynamic modeling of post-combustion carbon capture process based on multi-gate mixture-of-experts incorporating dual-stage attention-based encoder-decoder network

Solvent-based post-combustion carbon capture (PCC) technology is a promising, near-term solution for decarbonizing power generation and industrial facilities. Model-based process simulation is crucial for the optimal design and operation of the PCC process. Recently, data-driven models have gained attention due to their adaptability, efficient computation and high accuracy. However, the nonlinearity, strong couplings and multi-time scale features of the PCC process pose significant challenges for model identification. To this end, this paper proposes a multi-gate mixture-of-experts incorporating dual-stage attention-based encoder-decoder (MMoE-DAED) network for dynamic modeling of the PCC process under wide operating conditions. An encoder-decoder composed of long short-term memory (LSTM) network is employed to extract features from the time-dependent input data and learn the complex dynamic interactions caused by the inertial and delay properties of the process. Dual-stage attention mechanism is incorporated into the encoder and decoder respectively to select the most relevant input features and their correlations within the time series data. To enhance multi-output prediction accuracy, multi-gate mixture-of-experts (MMoE) framework that considers correlations of multitask learning is implemented. Simulation results using operating data from a PCC experimental setup indicate that the proposed modeling approach accurately predicts the steady-state values and dynamic trends of the CO2 capture rate and stripper bottom temperature over a wide operating range. The RMSE, MAPE and R2 indices for the CO2 capture rate are 2.1592, 0.0295, 0.9641, respectively, and for the stripper bottom temperature are 0.1491, 0.0003, 0.9833, respectively. Validations on a PCC simulator further verify the accuracy and efficiency of the MMoE-DAED model, which enables an 80.87% reduction in computation time compared to the simulator. This paper points to a new direction for the data-driven dynamic modeling of complex energy conversion processes.

WS-BiTM: Integrating White Shark Optimization with Bi-LSTM for Enhanced Autism Spectrum Disorder Diagnosis

Autism Spectrum Disorder (ASD) is a multifaceted neurodevelopmental condition marked by challenges in social communication, sensory processing, and behavioral regulation. The delayed diagnosis of ASD significantly impedes timely interventions, which can exacerbate symptom severity. With approximately 62 million individuals affected worldwide, the demand for efficient diagnostic tools is critical. This study introduces a novel framework that combines a White Shark Optimization (WSO)-based feature selection method with a Bidirectional Long Short-Term Memory (Bi-LSTM) classifier for enhanced autism classification. Utilizing the WSO technique, we identify key features from autism screening datasets, which markedly improves the model's predictive capabilities. The optimized feature set is then processed by the Bi-LSTM classifier, enhancing its efficiency in handling sequential data. We comprehensively address methodological challenges, including overfitting, generalization, interpretability, and computational efficiency. Furthermore, we conduct a comparative analysis against baseline algorithms such as Neural Networks, Convolutional Neural Networks (CNN), and Long Short-Term Memory (LSTM) networks, while also employing Particle Swarm Optimization (PSO) for feature selection validation. We evaluate performance metrics, including accuracy, F1-score, specificity, precision, and sensitivity across three ASD datasets: Toddlers, Adults, and Children. Our results demonstrate that the WS-BiTM model significantly outperforms baseline methods, achieving accuracies of 97.6%, 96.2%, and 96.4% on the respective datasets. Additionally, we implemented leave-one-dataset cross-validation and confirmed the statistical significance of our findings through a paired t-test, supplemented by an ablation study to detail the contributions of individual model components. These findings highlight the potential of the WS-BiTM model as a robust tool for ASD classification.

Enhancing the performance of runoff prediction in data-scarce hydrological domains using advanced transfer learning

Accurate hydrological predictions are often hindered by the lack of stream gauges in data-scarce regions, where traditional transfer learning (TL) models like Long Short-Term Memory (LSTM) networks often face limitations due to reduced accuracy and adaptability. To enhance runoff prediction in such regions, we developed DAformer, a novel TL approach that integrates domain adversarial neural networks with the Informer model. Trained on comprehensive runoff data from U.S. basins, DAformer was applied to three basins in Chile and the Chaersen basin in China, demonstrating an effective transfer from data-rich to data-scarce environments. Results show that DAformer significantly outperforms LSTM-based models, improving forecast accuracy by 16.1% for 1-day lead time and by 100.5% for 5-day lead time. These improvements indicate that the DAformer model not only enhances prediction accuracy but also holds substantial practical implications for flood risk management and water resource planning in regions with limited data availability. By clustering basins based on Shuttle Radar Topography Mission (SRTM) and other geographical data, we found that relying on multiple source basins further enhances the performance. DAformer, therefore, serves as a robust and scalable method for enhancing runoff prediction for regions with limited data.

An Efficient DDoS Attack Detection and Prevention Model using fusion Heuristic Enhancement of Deep Learning Approach in FANET Sector

Problem OverviewA Flying Ad-hoc Network (FANET) is a decentralized communication network formed by Unmanned Aerial Vehicles (UAVs). However, it faces significant security challenges due to its decentralized nature and mobility. One of the most critical threats is the Distributed Denial-of-Service (DDoS) attack, which aims to overcome the network by flooding it with malicious actions, disrupting communication, and causing system failures. The primary challenge lies in detecting and mitigating such attacks in real-time, given the dynamic and complex nature of FANETs, where the rapid movement of UAVs complicates monitoring and defending mechanisms. MethodologyA novel detection and prevention model has been proposed to address the issue of DDoS attacks utilizing a hybrid heuristic-based machine learning approach tailored for FANET environments. The model begins by collecting necessary information using benchmark datasets to train and enhance the detection performance. The collected data is used for detecting the DDoS attack using Hybrid Deep Learning (HDL). The HDL model developed by combining Deep Temporal Convolutional Networks (DTCN) and Long Short-Term Memory (LSTM) networks, offers a powerful solution. DTCNs excel at detecting short-term, rapidly changing patterns in network traffic, which is essential for recognizing the instant effects of topology changes in FANETs. At the same time, LSTMs are designed to handle long-term dependencies and can used to the evolving patterns of UAV movement and traffic behavior over time. Here, the hyper-parameters are optimized by proposing the Hybrid of Water Strider and Cuckoo Search (HWSCS). Water Strider optimization (WSA) and Cuckoo Search Optimizer (CSO) are combined in the HWSCS algorithm. This algorithm is a versatile and powerful tool for optimizing multifaceted systems across different fields. Its combination of local and global search abilities allows it to find optimal solutions in complicated environments, making it invaluable for tasks ranging from machine learning to network security and beyond. In network security, HWSCS is effective in optimizing the parameters of intrusion detection systems, particularly in detecting complex cyber threats like DDoS attacks, ensuring more accuracy while reducing false positives. After detecting the DDoS attack, it is prevented from communication by establishing the routing process. This routing process involves rerouting the network traffic away from the affected areas and towards secure servers, effectively isolating the attack and minimizing its impact on the network. Additionally, it allows for greater flexibility and adaptability in responding to evolving threats in realtime. Here, the attack mitigation is accomplished by finding the Optimal Link State Routing (OLSR), estimated by the hybrid HWSCS algorithm. This algorithm determines the most efficient path for redirecting traffic away from the targeted servers, preventing overload, and maintaining network stability. The integration of HWSCS with OLSR not only improves network security but also proves the importance of innovative solutions in protecting malicious activities. Lastly, the model's performance is validated and measured with different metrics. ResultsThe proposed model demonstrated superior performance compared to traditional models. It achieved a recall of 93.87, significantly higher than other approaches like 89.22 for MobileNet, 89.40 for DTCN, 88.44 for LSTM, and 88.35 for DTCN-LSTM. This improved detection accuracy, combined with the effective routing mechanism, ensures better prevention and mitigation of DDoS attacks in FANETs. The results confirm the model's ability to not only detect attacks but also minimize network disruption, providing a robust solution for maintaining secure and stable communication in FANET environments.

Combining residual convolutional LSTM with attention mechanisms for spatiotemporal forest cover prediction

Understanding spatiotemporal variations in forest cover is crucial for effective forest resource management. However, existing models often lack accuracy in simultaneously capturing temporal continuity and spatial correlation. To address this challenge, we developed ResConvLSTM-Att, a novel hybrid model integrating residual neural networks, Convolutional Long Short-Term Memory (ConvLSTM) networks, and attention mechanisms. We evaluated ResConvLSTM-Att against four deep learning models: LSTM, combined convolutional neural network and LSTM (CNN-LSTM), ConvLSTM, and ResConvLSTM for spatiotemporal prediction of forest cover in Tasmania, Australia. ResConvLSTM-Att achieved outstanding prediction performance, with an average root mean square error (RMSE) of 6.9% coverage and an impressive average coefficient of determination of 0.965. Compared with LSTM, CNN-LSTM, ConvLSTM, and ResConvLSTM, ResConvLSTM-Att achieved RMSE reductions of 31.2%, 43.0%, 10.1%, and 6.5%, respectively. Additionally, we quantified the impacts of explanatory variables on forest cover dynamics. Our work demonstrated the effectiveness of ResConvLSTM-Att in spatiotemporal data modelling and prediction.

BCBA: An IIoT encrypted traffic classifier based on a serial network model

With the rapid development of the Industrial Internet of Things (IIoT), ensuring the security and privacy of network traffic has become particularly important. Classifying and identifying encrypted traffic is a critical step in enhancing network security, but traditional traffic classification methods often struggle to handle the complexities of the IIoT environment. In this paper, we propose the BCBA model, a bidirectional encoder representation from transformers (BERT)-based serial network model, which significantly improves the encrypted traffic classification performance and can be trained more than four times faster than the original BERT classifier. The BCBA method obtains word vector representations from the embedding layer of the pretrained BERT model, uses convolutional neural networks (CNNs) to extract local features, employs bidirectional long short-term memory (BiLSTM) networks to capture temporal dependencies in traffic data, and leverages a multihead self-attention mechanism to improve global dependency understanding. In experiments on six types of regular encrypted traffic from the ISCXVPN2016 dataset, the BCBA model achieved an F1 score of 99.10%, outperforming traditional traffic classification techniques across multiple performance metrics. This study demonstrates the effectiveness of deep learning in enhancing the security of the IIoT. It also provides new perspectives and technical routes for future research, particularly in encrypted traffic processing and classification applications.

ROV localization in the nuclear reactor pressure vessel using LSTM and an improved adaptive Kalman Filter

In this paper, a remotely operated vehicle (ROV) localization algorithm for nuclear reactor pressure vessels is introduced. Traditional machine vision localization suffers from visual blind zones. To address this issue, a long short-term memory (LSTM) network was employed to obtain the ROV’s position. To further enhance the accuracy, an improved Sage-Husa adaptive Kalman Filter (SHAKF) was applied to eliminate cumulative errors from LSTM. Finally, upper-diagonal decomposition and confidence checks were used to enhance the traditional SHAKF, addressing issues associated with the non-positive definite error covariance matrix and sensor limitations. Simulated and experimental results demonstrated that the LSTM network significantly reduced the cumulative error from trajectory projection localization. The root mean square error of the localization results was reduced from 51.776 to 12.808 mm due to the improvements made to the traditional SHAKF. Additionally, the improved SHAKF effectively minimized the cumulative LSTM localization error.

XSShield: A Novel Dataset and Lightweight Hybrid Deep Learning Model for XSS Attack Detection

With the proliferation of web applications, cross-site scripting (XSS) attacks have increased significantly and now pose a significant threat to users' information security and privacy. To enhance the efficiency of XSS attack detection, the adoption of machine learning (ML) and deep learning (DL) techniques offers promising solutions, but their effectiveness is limited by the lack of comprehensive and diverse datasets. Moreover, existing approaches often prioritize detection accuracy over real-time processing capabilities, which are essential for effective defense. To address these challenges, in this paper, we propose a novel framework that automatically collects web resources, efficiently extracts informative features, and constructs an up-to-date XSS attack dataset, which is then used to train a machine learning-based XSS detection model. Using this framework, we created and published a well-structured dataset over 100,000 samples for the research community. Furthermore, we present a hybrid detection model that leverages the strengths of both Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks. Extensive evaluations of our dataset demonstrate that the proposed model outperforms other baseline ML models across various metrics, including processing rate. Notably, our model achieves an accuracy of 99.27% while maintaining a low false positive rate of 0.06% and high processing rate of exceeding 1000 samples per second. These results highlight its high accuracy and robustness in detecting XSS, and suitability for real-time applications. Our work presents a comprehensive solution for enhancing web application security by providing a diverse dataset and a high-accuracy detection model with low latency.

Deep learning method for predicting the complex nonlinear dynamics of passively mode-locked fiber laser

The dynamic evolution processes are highly complicated nonlinear dynamic processes in passively mode-locked fiber laser systems. Here, an artificial intelligence (AI) model is employed to predict the complex dynamic processes, which uses the long short-term memory network method, serving as an alternative to the numerical calculation of the nonlinear Schrödinger equation (NLSE). We specifically emphasize the complex evolution processes under different gain saturation energies, comparing the results predicted by the AI model with those simulated by the NLSE. The predicted results of the AI model are in good agreement with the simulated results of NLSE. The root mean square errors of test samples in this study are all below 0.15. Furthermore, with GPU acceleration, the AI model achieves a mean simulation time of 0.452 s for 6000 roundtrips, approximately 2391 times faster than the numerical solution of NLSE executed on a CPU.

Particle swarm optimization tuned multi-headed long short-term memory networks approach for fuel prices forecasting

Increasing global energy demands and decreasing stocks of fossil fuels have led to a resurgence of research into energy forecasting. Artificial intelligence, explicitly time series forecasting holds great potential to improve predictions of cost and demand with many lucrative applications across several fields. Many factors influence prices on a global scale, from socio-economic factors to distribution, availability, and international policy. Also, various factors need to be considered in order to make an accurate forecast. By analyzing the current literature, a gap for improvements within this domain exists. Therefore, this work suggests and explores the potential of multi-headed long short-term memory models for gasoline price forecasting, since this issue was not tackled with multi-headed models before. Additionally, since the computational requirements for such models are relatively high, work focuses on lightweight approaches that consist of a relatively low number of neurons per layer, trained in a small number of epochs. However, as algorithm performance can be heavily dependent on appropriate hyper-parameter selections, a modified variant of the particle swarm optimization algorithm is also set forth to help in optimizing the model’s architecture and training parameters. A comparative analysis is conducted using energy data collected from multiple public sources between several contemporary optimizers. The outcomes are put through a meticulous statistical validation to ascertain the significance of the findings. The best-constructed models attained a mean square error of just 0.044025 with an R-squared of 0.911797, suggesting potential for real-world use.