Long Short-term Memory Network Research Articles

ActiWise: Insight on Human Activity Recognition Using Deep Learning Approaches

In this study, we investigate the fusion of Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks for human activity recognition (HAR). By integrating hierarchical spatial features extracted by CNNs with LSTM networks' temporal modelling capabilities, our approach excels in discerning nuanced patterns from raw sensor data collected via wearable devices. Through rigorous experimentation and validation, our CNN+LSTM model demonstrates robust performance in accurately classifying a spectrum of human activities. This research advances HAR methodologies, shedding light on the synergistic interplay between spatial and temporal modelling in activity recognition, with implications across healthcare, sports analytics, and human-computer interaction domains. Index Terms- Human activity recognition, deep learning, Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM).

Predictive Analysis of COVID-19 Epidemic in Thailand: Evaluating Control Lockdown Measures using LSTM Networks

This study addresses the critical objective of evaluating the effectiveness of non-pharmaceutical lockdown measures implemented during COVID-19 outbreaks in Thailand. Assessing the outcome of these measures provides valuable insight that can inform and guide response to future outbreaks. Utilizing a closed-loop forecasting model built on Long Short-Term Memory (LSTM) networks, the research focuses on achieving precise daily forecasts of COVID-19 cases. The methodology involves optimizing hyperparameters through grid-search and incorporating training data from other countries that implemented similar measures. The LSTM, configured with an optimal number of hidden processing units, utilizes past lagged data of daily infected cases as predictors to generate multi-step-ahead predicted values, which are subsequently used as predictors in a recursive approach. As a result, the predicted cases closely align with measured data, facilitating the estimation of the effective reproduction number (Reff) to assess the performance of lockdown measures. The effectiveness of the lockdown measures is quantified at different time intervals: 51%, 41%, and 23% one day after implementation, increasing to 84%, 98%, and 34% after one week, and reaching 96%, 99%, and 73% at the endpoint of the first, second, and fourth waves of infection, respectively. Throughout these waves, the final Reff remains below 1, indicating ongoing but controllable COVID-19, demonstrating the efficacy of the implemented lockdown measures. It is noted that these results are based on specific LSTM model, as the effectiveness of lockdown measures may vary with alternative modeling approaches. Therefore, the findings should be interpreted in the context of this LSTM-framework.

Deep learning neural networks with input processing for vibration-based bearing fault diagnosis under imbalanced data conditions

Deep learning (DL) networks, such as convolutional neural networks (CNNs) and long short-term memory (LSTM), have gained popularity for bearing fault diagnosis utilizing raw vibration signals. However, their accuracy and stability are compromised when facing imbalanced real-world datasets. This research investigates the impact of imbalanced datasets and explores the potential of signal processing techniques on network inputs compared to the direct use of raw vibration signals. The DL techniques studied include LSTM, one-dimensional CNN, and two-dimensional (2D) CNN, and a novel hybrid 2DCNNLSTM algorithm, incorporating signal processing methods such as Fourier transform and continuous wavelet transform while maintaining nearly equal parameters and the same base architecture. The proposed hybrid 2DCNNLSTM algorithm combines the strengths of LSTM and CNN, allowing for improved bearing diagnosis by capturing both spatial and temporal information in vibration signals. The proposed 2DCNNLSTM algorithm also considers multi-channel input augmenting raw vibration signal, mean, and variance channels to extract meaningful features and enhance classification efficiency. The publicly available Case Western Reserve University benchmark-bearing test rig dataset with ten fault classes, the Paderborn University dataset with three fault classes, and NASA Centre for Intelligent Maintenance Systems bearing datasets with five fault classes are utilized to test the proposed deep learning networks’ accuracy, effectiveness, robustness, and stability. The studies reveal that the hybrid 2DCNNLSTM-based networks outperform both CNN and LSTM networks, even without input processing. Further, utilizing multi-channel input by augmenting the 2D raw signal with mean and variance value channels proves to be more efficient in handling imbalanced and complex datasets while employing a 2DCNNLSTM-based network.

Short-term photovoltaic energy generation for solar powered high efficiency irrigation systems using LSTM with Spatio-temporal attention mechanism

Solar irrigation systems should become more practical and efficient as technology advances. Automation and AI-based technologies can optimize solar energy use for irrigation while reducing environmental impacts and costs. These innovations have the potential to make agriculture more environmentally friendly and sustainable. Solar irrigation system implementation can be hampered by a lack of technical expertise in installation, operation, and maintenance. It must be technically and economically feasible to be practical and continuous. Due to weather and solar irradiation, photovoltaic power generation is difficult for high-efficiency irrigation systems. As a result, more precise photovoltaic output calculations could improve solar power systems. Customers should benefit from increased power plant versatility and high-quality electricity. As a result, an artificial intelligence-powered automated irrigation power-generation system may improve the existing efficiency. To predict high-efficiency irrigation system power outputs, this study proposed a spatial and temporal attention block-based long-short-term memory (LSTM) model. Using MSE, RMSE, and MAE, the results have been compared to pre-existing ML and a simple LSTM network. Moreover, it has been found that our model outperformed cutting-edge methods. MAPE was improved by 6–7% by increasing Look Back (LB) and Look Forward (LF). Future goals include adapting the technology for wind power production and improving the proposed model to harness customer behavior to improve forecasting accuracy.

Prediction of Parking Space Availability Using Improved MAT-LSTM Network

Prediction of Parking Space Availability Using Improved MAT-LSTM Network

A novel solar irradiance forecasting method based on multi-physical process of atmosphere optics and LSTM-BP model

Prediction of solar irradiance is crucial for minimizing energy costs and ensuring high power quality in electrical power grids that incorporate distributed solar photovoltaic generation. Traditional methods have focused mainly on time series prediction of irradiance or studies of the relationship between irradiance and environmental factors at a given moment. This paper presents a novel global horizontal irradiance forecasting method based on multi-physical process of atmosphere optics and LSTM-BP model. In the proposed method, cloud fraction, cloud albedo and aerosol optical thickness are selected as critical characteristic factors in the sunlight-atmosphere interaction. Long-short term memory (LSTM) is utilized to extract temporal features of the critical characteristic factors in various physical processes of atmospheric optics and predict their values. Subsequently, back propagation (BP) neural network is constructed to investigate the relationship between these characteristic factors and global horizontal irradiance under a given point in time. The global horizontal irradiance is predicted by combining the predicted values of critical characteristic factors and BP neural network for modelling irradiance decay process. To demonstrate the superior performance of the proposed model, the results obtained by the proposed method are compared with those of the LSTM network model for time series forecasting, in which six error evaluation indicators are used in the comparison. The proposed method for solar irradiance forecasting is verified in different time scales and under various weather conditions and shows better accuracy comparing with time series forecasting methods.

Adaptive Kalman filter with LSTM network assistance for abnormal measurements

A classic state estimation method, the Kalman filter integrates prior information, system dynamics models, and measurement data to achieve posterior state estimations. However, measurements often encounter various unknown disturbances, leading to abnormal or inaccurate measurements and subsequently impacting the performance of Kalman filtering. In response to this challenge, this paper introduces a novel adaptive Kalman filter approach aided by posterior state estimations using the Long short-term memory (LSTM) networks. The proposed method begins by utilizing prior residuals to construct a chi-square distribution model, which facilitates the detection of abnormal measurements within data. Upon identifying abnormal measurements, an LSTM network is employed to generate alternative predictive measurements, replacing the original inaccurate measurement. This approach enhances the model’s capability to handle complex relationship and measurement uncertainties and improves filtering estimation accuracy. A noise covariance adjustment method is introduced in extreme cases where alternative predictive measurements alone are insufficient for filter requirements. This method mitigates the adverse effects of abnormal measurements on posterior estimation. Throughout the entire adaptive assistance process involving the LSTM network, deep learning maintains a stable structure of the filter while enhancing adaptability. This strategy ensures the filter’s resilience in dynamic environments with unknown factors by providing predicted measurements conforming to the chi-square distribution and corresponding measurement noise covariance. Ultimately, the efficacy of the proposed algorithm is validated through simulations and experiments involving vehicle positioning with inertial sensors.

Mixed Learning- and Model-Based Mass Estimation of Heavy Vehicles

This research utilized long short-term memory (LSTM) to oversee an RLS-based mass estimator based on longitudinal vehicle dynamics for heavy-duty vehicles (HDVs) instead of using the predefined rules. A multilayer LSTM network that analyzed parameters such as vehicle speed, longitudinal acceleration, engine torque, engine speed, and estimated mass from the RLS mass estimator was employed as the supervision method. The supervisory LSTM network was trained offline to recognize when the vehicle was operated so that the RLS estimator gave an estimate with the desired accuracy and the network was used as a reliability flag. High-fidelity simulation software was employed to collect data used to train and test the network. A threshold on the error percentage of the RLS mass estimator was used by the network to check the reliability of the algorithm. The preliminary findings indicate that the reliability of the RLS mass estimator could be predicted by using the LSTM network.

Deep learning-based modeling of the cyclic behavior of replaceable fuse buckling-restrained braces (BRBs)

This paper presents a novel approach for predicting the behavior of buckling restrained braces (BRBs) using long short-term memory (LSTM) and residual neural network (ResNet) models. Nonlinear finite element analysis (NLFEA) usually requires 3D modeling of complex parts and assigning various material and simulation properties before conducting the analysis, which can be time-consuming and computationally costly. LSTM networks are particularly well-suited for modeling sequential data, such as time series, and ResNet is a discriminative deep learning-based model that can handle deeper architectures without the issue of vanishing gradient. Developing a general LSTM model requires a comprehensive database for training, cross-validation, and blind testing. Even if such a database were available, concatenating data from different BRB specimens would not be feasible when modeling a data series. Such a method would intertwine the features of the individual specimens, rendering the data useless for machine-learning purposes. On the other hand, a well-trained LSTM can predict cyclic behavior for specimens that resemble the training data. As such, a group of well-trained models can produce superior predictions when used on individually familiar data. Hence, there is a need for a prescreening classification stage. The framework presented herein utilizes an AI classifier (ResNet) that automatically recognizes the BRB specimen type of the input data and directs it to the appropriate LSTM model for prediction. The presented framework demonstrates excellent predictions with an accuracy greater than 99%. It is customizable, adaptable, and scalable to include different BRB specimens. The input data consisted of the displacement of the BRB specimens based on the AISC 341–10 qualification recommendations, while the output was the corresponding force (hysteretic response to cyclic loading). The proposed framework can be valuable in earthquake engineering applications as it demonstrates superior predictions.

A survey on cryptocurrency price prediction

A type of digital currency known as a cryptocurrency allows all transactions to be completed online. There is no hard cash version of this soft currency. We highlight that a decentralized currency differs from a centralized currency in the any user of a virtual currency can purchase services without the need for third parties to get involved. Due to its extreme price volatility, using these cryptocurrencies has an impact on trade and international relations. Moreover, the constantly fluctuating oscillations indicate the urgent need for a more precise method of predicting this price. Deep learning techniques that use effective learning models for training data, including the LSTM, GRU, and Feedback Neural Network, can be used to do this. Benchmark datasets are used to test the suggested strategy. That brings us to the neural network, one of the clever data mining technologies that researchers in many domains have been using for the past ten years. In the current economy, stock market data is essential. There are two types of forecasting methodologies: nonlinear models (ARCH, GARCH, Neural Network) and linear models (AR, MA, ARIMA, ARMA). To forecast a company's stock price based on past prices, we employed the Box Jenkins Model also known as ARIMA, and Long Short-Term Memory (LSTM), and Feedback Neural Network also known as RNN.

Daily Streamflow Forecasting Using Networks of Real-Time Monitoring Stations and Hybrid Machine Learning Methods

Considering the increased risk of urban flooding and drought due to global climate change and rapid urbanization, the imperative for more accurate methods for streamflow forecasting has intensified. This study introduces a pioneering approach leveraging the available network of real-time monitoring stations and advanced machine learning algorithms that can accurately simulate spatial–temporal problems. The Spatio-Temporal Attention Gated Recurrent Unit (STA-GRU) model is renowned for its computational efficacy in forecasting streamflow events with a forecast horizon of 7 days. The novel integration of the groundwater level, precipitation, and river discharge as predictive variables offers a holistic view of the hydrological cycle, enhancing the model’s accuracy. Our findings reveal that for a 7-day forecasting period, the STA-GRU model demonstrates superior performance, with a notable improvement in mean absolute percentage error (MAPE) values and R-square (R2) alongside reductions in the root mean squared error (RMSE) and mean absolute error (MAE) metrics, underscoring the model’s generalizability and reliability. Comparative analysis with seven conventional deep learning models, including the Long Short-Term Memory (LSTM), the Convolutional Neural Network LSTM (CNNLSTM), the Convolutional LSTM (ConvLSTM), the Spatio-Temporal Attention LSTM (STA-LSTM), the Gated Recurrent Unit (GRU), the Convolutional Neural Network GRU (CNNGRU), and the STA-GRU, confirms the superior predictive power of the STA-LSTM and STA-GRU models when faced with long-term prediction. This research marks a significant shift towards an integrated network of real-time monitoring stations with advanced deep-learning algorithms for streamflow forecasting, emphasizing the importance of spatially and temporally encompassing streamflow variability within an urban watershed’s stream network.

Deepfake Face Detection Using Machine Learning with LSTM

Deep fake videos, which employ artificial intelligence to manipulate and generate highly convincing fake content, have emerged as a significant threat to society, potentially undermining trust in visual media. Detecting these deceptive videos is outmost importance to combat the spread of misinformation and protect the integrity of digital media. In this study, we propose a novel approach for deep fake face video detection utilizing Long Short-Term Memory (LSTM) networks, a type of Recurrent Neural Network (RNN). Our approach capitalizes on the temporal patterns and context within video sequences, harnessing the unique strengths of LSTM in capturing sequential information. We demonstrate the effectiveness of our methodology by training the LSTM network on a diverse dataset comprising both real and deep fake videos. The network’s ability to learn temporal dependencies and identify inconsistencies in facial expressions, eye movements, and other subtle cues allows it to distinguish between genuine and manipulated content. To further enhance the accuracy and robustness of our deep fake face detection system, we integrate pre-processing techniques for frame- level analysis, such as optical flow computation and facial landmarks extraction. Additionally, we employ a comprehensive ensemble of LSTM models and other machine learning algorithms to improve the overall detection performance. In our experiments, we evaluate the LSTM-based deep fake detection system on a large-scale dataset of both known and unseen deep fake videos, achieving high detection accuracy and low false positive rates. We also compare our approach with existing methods, demonstrating its superiority in terms of robustness and generalization. The results of this study signify the potential of LSTM-based models for mitigating the adverse effects of deep fake content on society. As deep fake technology continues to evolve, our approach showcases a promising step towards combating the dissemination of deceptive multimedia, promoting media integrity, and upholding trust in visual information. Keywords: LSTM Networks, Recurrent Neural Network, Optical flow computation, Facial landmarks extraction, False positive rates.

A Path-Planning Method for UAV Swarm under Multiple Environmental Threats

To weaken or avoid the impact of dynamic threats such as wind and extreme weather on the real-time path of a UAV swarm, a path-planning method based on improved long short-term memory (LSTM) network prediction parameters was constructed. First, models were constructed for wind, static threats, and dynamic threats during the flight of the drone. Then, it was found that atmospheric parameters are typical time series data with spatial correlation. The LSTM network was optimized and used to process time series parameters to construct a network for predicting atmospheric parameters. The state of the drone was adjusted in real time based on the prediction results to mitigate the impact of wind or avoid the threat of extreme weather. Finally, a path optimization method based on an improved LSTM network was constructed. Through simulation, it can be seen that compared to the path that does not consider atmospheric effects, the optimized path has significantly improved flightability and safety.

A Temperature Control Method of Lysozyme Fermentation Based on LRWOA-LSTM-PID

In order to overcome the difficulty of parameter tuning caused by the large lag and time-varying nonlinearity of the tank for lysozyme fermentation, a temperature control method based on LRWOA-LSTM-PID is proposed in this paper. Firstly, according to the intrinsic mechanism of the fermenter, a temperature mechanism model based on a dynamic equation is designed, which can better reflect the temperature changes in the fermenter. Secondly, a Proportional Integral Derivative (PID) parameter tuning method based on a Long-Short Term Memory Network (LSTM) is proposed, which takes advantage of the ability of LSTM to learn time sequence information and obtains the variation trend between error sequences under continuous time sampling, thereby adjusting network weights more reasonably and accelerating PID parameter tuning. Finally, a Whale Optimization Algorithm (WOA) based on the Lévy flight and random walk strategy (LRWOA) is proposed for the initialization of LSTM parameters; this algorithm has excellent optimization capabilities and overcomes the problem of LSTM falling into local optimal solutions prematurely during parameter randomization. The results show that the method proposed in this paper can achieve rapid tuning of PID parameters, thereby improving the convergence speed of the system and reducing system overshoot.

Demand forecasting for platelet usage: From univariate time series to multivariable models.

Platelet products are both expensive and have very short shelf lives. As usage rates for platelets are highly variable, the effective management of platelet demand and supply is very important yet challenging. The primary goal of this paper is to present an efficient forecasting model for platelet demand at Canadian Blood Services (CBS). To accomplish this goal, five different demand forecasting methods, ARIMA (Auto Regressive Integrated Moving Average), Prophet, lasso regression (least absolute shrinkage and selection operator), random forest, and LSTM (Long Short-Term Memory) networks are utilized and evaluated via a rolling window method. We use a large clinical dataset for a centralized blood distribution centre for four hospitals in Hamilton, Ontario, spanning from 2010 to 2018 and consisting of daily platelet transfusions along with information such as the product specifications, the recipients' characteristics, and the recipients' laboratory test results. This study is the first to utilize different methods from statistical time series models to data-driven regression and machine learning techniques for platelet transfusion using clinical predictors and with different amounts of data. We find that the multivariable approaches have the highest accuracy in general, however, if sufficient data are available, a simpler time series approach appears to be sufficient. We also comment on the approach to choose predictors for the multivariable models.

Analysis and Prediction of Urban Surface Transformation Based on Small Baseline Subset Interferometric Synthetic Aperture Radar and Sparrow Search Algorithm-Convolutional Neural Network-Long Short-Term Memory Model.

With the acceleration of urbanisation, urban areas are subject to the combined effects of the accumulation of various natural factors, such as changes in temperature leading to the thermal expansion or contraction of surface materials (rock, soil, etc.) and changes in precipitation and humidity leading to an increase in the self-weight of soil due to the infiltration of water along the cracks or pores in the ground. Therefore, the subsidence of urban areas has now become a serious geological disaster phenomenon. However, the use of traditional neural network prediction models has limitations when examining the causal relationships between time series surface deformation data and multiple influencing factors and when applying multiple influencing factors for predictive analyses. To this end, Sentinel-1A data from March 2017 to February 2023 were used as the data source in this paper, based on time series deformation data acquired using the small baseline subset interferometric synthetic aperture radar (SBAS-InSAR) technique. A sparrow search algorithm-convolutional neural network-long short-term memory (SSA-CNN-LSTM) neural network prediction model was built. The six factors of temperature, humidity, precipitation, and ground temperature at three different depths below the surface (5 cm, 10 cm, and 15 cm) were taken as the input of the model, and the surface deformation data were taken as the output of the neural network model. The correlation between the spatial and temporal evolution characteristics of the ground subsidence in urban areas and various influencing factors was analysed using grey correlation analysis, which proved that these six factors contribute to some extent to the deformation of the urban surface. The main urban area of Hohhot City, Inner Mongolia Autonomous Region, was used as the study area. In order to verify the efficacy of this neural network prediction model, the prediction effects of the multilayer perceptron (MLP), backpropagation (BP), and SSA-CNN-LSTM models were compared and analysed, with the values of the correlation coefficients of the feature points of A1, B1, and C1 being in the range of 0.92, 0.83, and 0.93, respectively. The results show that compared with the traditional MLP and BP neural network models, the SSA-CNN-LSTM model achieves a higher performance in predicting time series surface deformation data in urban areas, which provides new ideas and methods for this area of research.

Machine Learning Applied to Electron Beam Lithography to Accelerate Process Optimization of a Contact Hole Layer.

Determining the lithographic process conditions with high-resolution patterning plays a crucial role in accelerating chip manufacturing. However, lithography imaging is an extremely complex nonlinear system, and obtaining suitable process conditions requires extensive experimental attempts. This severely creates a bottleneck in optimizing and controlling the lithographic process conditions. Herein, we report a process optimization solution for a contact layer of metal oxide nanoparticle photoresists by combining electron beam lithography (EBL) experiments with machine learning. In this solution, a long short-term memory (LSTM) network and a support vector machine (SVM) model are used to establish the contact hole imaging and process condition classification models, respectively. By combining SVM with the LSTM network, the process conditions that simultaneously satisfy the requirements of the contact hole width and local critical dimension uniformity tolerance can be screened. The verification results demonstrate that the horizontal and vertical contact widths predicted by the LSTM network are highly consistent with the EBL experimental results, and the classification model shows good accuracy, providing a reference for process optimization of a contact layer.

Evaluation of streamflow predictions from LSTM models in water- and energy-limited regions in the United States

The application of Long Short-Term Memory (LSTM) models for streamflow predictions has been an area of rapid development, supported by advancements in computing technology, increasing availability of spatiotemporal data, and availability of historical data that allows for training data-driven LSTM models. Several studies have focused on improving the performance of LSTM models; however, few studies have assessed the applicability of these LSTM models across different hydroclimate regions. This study investigated the single-basin trained local (one model for each basin), multi-basin trained regional (one model for one region), and grand (one model for several regions) models for predicting daily streamflow in water-limited Great Basin (18 basins) and energy-limited New England (27 basins) regions in the United States using the CAMELS (Catchment Attributes and Meteorology for Large-sample Studies) data set. The results show a general pattern of higher accuracy in daily streamflow predictions from the regional model when compared to local or grand models for most basins in the New England region. For the Great Basin region, local models provided smaller errors for most basins and substantially lower for those basins with relatively larger errors from the regional and grand models. The evaluation of one-layer and three-layer LSTM network architectures trained with 1-day lag information indicates that the addition of model complexity by increasing the number of layers may not necessarily increase the model skill for improving streamflow predictions. Findings from our study highlight the strengths and limitations of LSTM models across contrasting hydroclimate regions in the United States, which could be useful for local and regional scale decisions using standalone or potential integration of data-driven LSTM models with physics-based hydrological models.

Anomaly Detection of Permanent Magnet Synchronous Motor Based on Improved DWT-CNN Multi-Current Fusion.

The Permanent Magnet Synchronous Motor (PMSM) is the power source maintaining the stable and efficient operation of various pieces of equipment; hence, its reliability is crucial to the safety of public equipment. Convolutional Neural Network (CNN) models face challenges in extracting features from PMSM current data. A new Discrete Wavelet Transform Convolutional Neural Networks (DW-CNN) feature with fusion weight updating Long Short-Term Memory (LSTM) anomaly detection is proposed in this paper. This approach combines Discrete Wavelet Transform (DWT) with high and low-frequency separation processing and LSTM. The anomaly detection method adopts DWT and CNN by separating high and low-frequency processing. Moreover, this method combines the hybrid attention mechanism to extract the multi-current signal features and detects anomalies based on weight updating the LSTM network. Experiments on the motor bearing real fault dataset and the PMSM stator fault dataset prove the method's strong capability in fusing current features and detecting anomalies.

LSTM Short-Term Wind Power Prediction Method Based on Data Preprocessing and Variational Modal Decomposition for Soft Sensors.

Soft sensors have been extensively utilized to approximate real-time power prediction in wind power generation, which is challenging to measure instantaneously. The short-term forecast of wind power aims at providing a reference for the dispatch of the intraday power grid. This study proposes a soft sensor model based on the Long Short-Term Memory (LSTM) network by combining data preprocessing with Variational Modal Decomposition (VMD) to improve wind power prediction accuracy. It does so by adopting the isolation forest algorithm for anomaly detection of the original wind power series and processing the missing data by multiple imputation. Based on the process data samples, VMD technology is used to achieve power data decomposition and noise reduction. The LSTM network is introduced to predict each modal component separately, and further sum reconstructs the prediction results of each component to complete the wind power prediction. From the experimental results, it can be seen that the LSTM network which uses an Adam optimizing algorithm has better convergence accuracy. The VMD method exhibited superior decomposition outcomes due to its inherent Wiener filter capabilities, which effectively mitigate noise and forestall modal aliasing. The Mean Absolute Percentage Error (MAPE) was reduced by 9.3508%, which indicates that the LSTM network combined with the VMD method has better prediction accuracy.