Convolutional Long Short-term Memory Research Articles

In-Situ Estimation of Bolt Clamping Force by Utilization of Its Effects on Flexural Vibration Propagation

After a bolt assembling process, non-destructive identification of the clamping force is important to minimize unexpected failure of the assembled joint structures. For in-situ inspection for every parts, a detachable device which allows generation and detection of vibration propagation was proposed in this study. Tone burst excitation was transmitted through the bolted-joint structure. The vibration propagation characteristics depended on the clamping force. With an increase in the clamping force, the surface vibration propagation speed increased. The vibration magnitude decreased due to increasing constraint. A band-pass filter was used to remove noise component. The filtered vibration propagation responses were used in convolutional long short-term memory (CLSTM) training. The performance of CLSTM model was verified through comparison with actual clamping force. With minimal physical influence on the assembled parts, the clamping force of bolted-joint was measured using the surface vibration propagation characteristics.

Al-Biruni Based Optimization of Rainfall Forecasting in Ethiopia

Rainfall plays a significant role in managing the water level in the reservoir. The unpredictable amount of rainfall due to the climate change can cause either overflow or dry in the reservoir. Many individuals, especially those in the agricultural sector, rely on rain forecasts. Forecasting rainfall is challenging because of the changing nature of the weather. The area of Jimma in southwest Oromia, Ethiopia is the subject of this research, which aims to develop a rainfall forecasting model. To estimate Jimma’s daily rainfall, we propose a novel approach based on optimizing the parameters of long short-term memory (LSTM) using Al-Biruni earth radius (BER) optimization algorithm for boosting the forecasting accuracy. Nash–Sutcliffe model efficiency (NSE), mean square error (MSE), root MSE (RMSE), mean absolute error (MAE), and R2 were all used in the conducted experiments to assess the proposed approach, with final scores of (0.61), (430.81), (19.12), and (11.09), respectively. Moreover, we compared the proposed model to current machine-learning regression models; such as non-optimized LSTM, bidirectional LSTM (BiLSTM), gated recurrent unit (GRU), and convolutional LSTM (ConvLSTM). It was found that the proposed approach achieved the lowest RMSE of (19.12). In addition, the experimental results show that the proposed model has R2 with a value outperforming the other models, which confirms the superiority of the proposed approach. On the other hand, a statistical analysis is performed to measure the significance and stability of the proposed approach and the recorded results proved the expected performance.

A Spatiotemporal Model for Global Earthquake Prediction Based on Convolutional LSTM

A Spatiotemporal Model for Global Earthquake Prediction Based on Convolutional LSTM

Evaluation of Deep Learning Models for Continuous Authentication Using Behavioral Biometrics

The traditional method to authenticate users on mobile devices or applications requires usernames and passwords. This approach authenticates the user only at the entry point of an application without providing continuous authentication during the whole session. This paper explores the use of behavioral biometrics, which involves tracking the unique movements of a user while interacting with a device, for continuous authentication on a mobile banking application. As a methodology, we use binary classification and explore the performance of deep learning algorithms. A dataset is collected from 45 participants using a mobile banking application in Turkey. We train four different types of deep architectures, including Multilayer Perceptron (MLP), LSTM, bi-directional LSTM, and convolutional LSTM. The dataset includes data from both touch screens and motion sensors. The results of the experiments reveal that MLP and the convolutional-LSTM algorithms achieve the best performance on raw data from both motion sensors and touch screens. Accuracy rates are over 99.85%, and FAR, FRR, and EER are below 0.5%.

Video anomaly detection by fusing self-attention and autoencoder

Video anomaly detection by fusing self-attention and autoencoder

Electricity price forecast based on the STL-TCN-NBEATS model

Taking long-term high-frequency electricity price data as the research content, this paper proposes seasonal and trend decomposition using loess-temporal convolutional network-neural basis expansion analysis for an interpretable time series forecasting (STL-TCN-NBEATS) model to solve the problems of low forecast accuracy caused by high volatility, high frequency and nonlinearity and poor interpretability of the deep learning model. By comparing the forecast effects of the temporal convolutional network-long short-term memory (TCN-LSTM), LSTM and other models, the main conclusions are as follows: (1) The hybrid model, STL-TCN-NBEATS, selected in this paper can effectively solve the problem of low forecast accuracy after reasonable selection of model parameters. The evaluation indexes of the root mean square error (RMSE) and the mean absolute percentage error (MAPE) were 3.7441 and 4.5044, respectively, which were 3.1416 and 2.1336 lower than those of the second-best model (TCN-LSTM). Compared with the autoregressive integrated moving average model (ARIMA), the accuracy was improved by approximately 49.18% (RMSE) and 60.35% (MAPE). (2) The STL-TCN-NBEATS model has better feature extraction ability, so it can obtain higher forecast accuracy. Since the construction of the TCN introduces extended causal convolution and residual blocks, the deep learning network model has better processing ability and robustness for large sample time series. Moreover, the NBEATS network structure enables the model to be trained quickly, and the experimental results verify the effectiveness and high accuracy of this method. (3) The model not only has high precision but also has some interpretability. By decomposing time series data into a trend term, period term and remainder term, the NBEATS and the TCN are used to process the trend term, period term and remainder term, respectively, so that the hybrid model can forecast electricity prices according to the traditional time series processing mode.

Fine-Grained Multivariate Time Series Anomaly Detection in IoT

Sensors produce a large amount of multivariate time series data to record the states of Internet of Things (IoT) systems. Multivariate time series timestamp anomaly detection (TSAD) can identify timestamps of attacks and malfunctions. However, it is necessary to determine which sensor or indicator is abnormal to facilitate a more detailed diagnosis, a process referred to as fine-grained anomaly detection (FGAD). Although further FGAD can be extended based on TSAD methods, existing works do not provide a quantitative evaluation, and the performance is unknown. Therefore, to tackle the FGAD problem, this paper first verifies that the TSAD methods achieve low performance when applied to the FGAD task directly because of the excessive fusion of features and the ignoring of the relationship’s dynamic changes between indicators. Accordingly, this paper proposes a multivariate time series fine-grained anomaly detection (MFGAD) framework. To avoid excessive fusion of features, MFGAD constructs two sub-models to independently identify the abnormal timestamp and abnormal indicator instead of a single model and then combines the two kinds of abnormal results to detect the fine-grained anomaly. Based on this framework, an algorithm based on Graph Attention Neural Network (GAT) and Attention Convolutional Long-Short Term Memory (A-ConvLSTM) is proposed, in which GAT learns temporal features of multiple indicators to detect abnormal timestamps and A-ConvLSTM captures the dynamic relationship between indicators to identify abnormal indicators. Extensive simulations on a real-world dataset demonstrate that the proposed algorithm can achieve a higher F1 score and hit rate than the extension of existing TSAD methods with the benefit of two independent sub-models for timestamp and indicator detection.

ConvSNow: A tailored Conv-LSTM architecture for weather nowcasting based on satellite imagery

Nowcasting represents a short-term weather forecast of how the atmospheric state will evolve during the next time period, typically less than two hours. It is vital for generating society-level emergency alerts in order to take timely actions and responses to potential disasters. The objective of the paper is to improve upon current nowcasting methods by applying a Deep Learning model that uses Convolutional Long-Short Term Memory Networks on a combination of satellite data. It is proposed a model ConvS Now for short-term prediction of satellite images that would be useful for precipitation nowcasting. The proposed model was trained and evaluated on satellite imagery collected by EUMESAT's Meteosat-11 satellite utilizing the Severe Storms RGB product. The experimental results performed a subset of the Meteosat-11 data spanning Europe demonstrate that this model can enhance weather short-term forecasting, reduce costs and time, and improve the general quality of predictions, as a normalized mean of absolute errors of 1.6% was attained, outperforming every other baseline approaches considered for comparison. A relative improvement of more than 30% has been achieved by the ConvS Now compared to the baselines, our proposed model being able to capture the spatio-temporal features of the weather evolution.

Convolutional LSTM Network for Real-Time Impulsive Sound Detection and Classification in Urban Environments

Convolutional LSTM Network for Real-Time Impulsive Sound Detection and Classification in Urban Environments

Convolutional LSTM Network for Heart Disease Diagnosis on Electrocardiograms

Convolutional LSTM Network for Heart Disease Diagnosis on Electrocardiograms

Retracted: Classification of Alzheimer's Disease Using Gaussian-Based Bayesian Parameter Optimization for Deep Convolutional LSTM Network.

[This retracts the article DOI: 10.1155/2021/4186666.].

Multispectral Crop Yield Prediction Using 3D-Convolutional Neural Networks and Attention Convolutional LSTM Approaches

In recent years, national economies are highly affected by crop yield predictions. By early prediction, the market price can be predicted, importing, and exporting plan can be provided, social, and economic effects of waste products can be minimized, and a program can be presented for humanitarian food aid. In addition, agricultural fields are constantly growing to generate products required. The use of machine learning (ML) methods in this sector can lead to the efficient production and high-quality agricultural products. Traditional predictive machine models were unable to find nonlinear relationships between data. Recently, there has been a revolution in prediction systems via the advancement of ML, which can be used to achieve highly accurate decision-making networks. Thus far, many strategies have been used to evaluate agricultural products, such as DeepYield, CNN-LSTM, and ConvLSTM. However, preferable prediction accuracy is required. In this study, two architectures have been proposed. The first model includes 2D-CNN, skip connections, and LSTM-Attentions. The second model comprises 3D-CNN, skip connections, and ConvLSTM Attention. The Input data given from MODIS products such as Land-Cover, Surface-Temperature, and MODIS-Land-surface from 2003 to 2018 on the county level over 1800 counties, where soybean is mainly cultivated in the USA. The proposed methods have been compared with the most recent models. Then, the results showed that the second proposed method notably outperformed the other techniques. In case of MAE, the second proposed method, DeepYield, ConvLSTM, 3DCNN, and CNN-LSTM obtained 4.3, 6.003, 6.05, 6.3, and 7.002, respectively.

Spatio-Temporal Human Action Recognition Model using Deep Learning Techniques

Two-stream human recognition achieved great success in the development of video action recognition using deep learning. Recently many studies have shown that two-stream action recognition is a powerful feature extractor. The main contribution in this work is to develop a two-stream model based on spatial and temporal networks using convolutional neural networks with a convolution long-short term memory. The two-stream model with ImageNet pre-trained weights is used to retrieve spatial and temporal features. Output feature maps of the two-stream model are fused using sum fusion and fed as input to convolutional long-short-term memory. SoftMax function is used to get the final classification score. To avoid overfitting, we have adopted the data augmentation techniques. Finally, we demonstrated that the proposed model performs well in comparison to state-of-the-art two-stream models with an accuracy of 96.1% on UCF 101 dataset and 70.9% accuracy on the HMDB dataset.

A Dual-Attention-Mechanism Multi-Channel Convolutional LSTM for Short-Term Wind Speed Prediction

Accurate wind speed prediction plays a crucial role in wind power generation and disaster avoidance. However, stochasticity and instability increase the difficulty of wind speed prediction. In this study, we proposed a dual-attention mechanism multi-channel convolutional LSTM (DACLSTM), collected European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis v5 (ERA5) near-ground element-grid data from some parts of North China, and selected elements with high correlations with wind speed to form multiple channels. We used a convolutional network for the feature extraction of spatial information, a Long Short-Term Memory (LSTM) network for the feature extraction of time-series information, and used channel attention with spatial attention for feature extraction. The experimental results show that the DACLSTM model can improve the accuracy of six-hour lead time wind speed prediction relative to the traditional ConvLSTM model and fully connected network long short-term memory (FC_LSTM).

Atrial fibrillation classification and detection from ECG recordings

ObjectiveAtrial fibrillation (AF) heart rhythm disorder is investigated under two topics: Persistent AF (PeAF) and Paroxysmal AF (PAF). Diagnosis and detection of PeAF is relatively easier than PAF and PAF generally remains unrecognized. It is observed that a significant number of studies in the literature focused on detection of AF. MethodsIn this study, four different approaches are examined for AF detection. The first one is based upon spectral features obtained from windowed ECG signals. In the second approach, distances between successor R peaks are used as features. Then in the third approach, P waves are detected from the ECG signals by using R peak positions and then the model is trained by those P waves. In those three approaches a deep learning architecture with bidirectional long short-term memory (BiLSTM) network is used. Finally, in the fourth approach, a convolutional long short-term memory (CLSTM) model with convolution and LSTM layers is used for classification. The data set used in this work is obtained from 4th China Physiological Signal Challenge-2021. ResultsAs the result of experimental studies, it is seen that classification approach based on spectral features provided the best training accuracy (0.9788) and classification based on P wave detection provided the best test accuracy (0.8765). SignificanceThis study compares PeAF and PAF detection and classification methods based on deep learning models using different approaches. BiLSTM networks being capable of reflecting time sensitive features of ECG, appeared to be superior to CNN and LSTM cascades.

Performance and complexity study of a neural network post-equalizer in a 638-nm laser transmission system through over 100-m plastic optical fiber

In recent years, plastic optical fiber (POF) has been considered as a promising cost-effective scheme for short-distance data communications, multimedia communication in cars, and in-house networks. However, due to the intrinsic nature of the relatively large numerical aperture of POF and the high attenuation rate, implementing high data rates over 100-m POF transmission length will be a significant challenge. We propose a scheme of high-speed 100-m POF transmission system based on a visible red laser and a cascaded neural network (NN) post-equalizer. To mitigate the nonlinear distortion induced by the POF, three different NNs, i.e., convolutional NN (CNN), long and short-term memory NN (LSTM), and cascaded NN structure consisting of convolutional layers and LSTM (CNN-LSTM), are employed as the post-equalizer. Experimental results show that using three different post-equalizers can significantly improve the system performance compared with the Volterra equalizer baseline. Among them, CNN-LSTM can outperform the others in terms of the bit error rate (BER) and the system Q-factor in the low nonlinear region. When the system operating in strong nonlinear region, CNN can achieve optimal performance at a lower system overhead of complexity. Finally, we successfully demonstrated a 100-m POF transmission system using 16 quadrature amplitude modulation discrete Fourier transform-spread orthogonal frequency division multiplexing modulation format at 1.8 Gbps with BER below 3.8 × 10 − 3 by utilizing CNN-LSTM.

Forecasting Steel Production in the World—Assessments Based on Shallow and Deep Neural Networks

Forecasting algorithms have been used to support decision making in companies, and it is necessary to apply approaches that facilitate a good forecasting result. The present paper describes assessments based on a combination of different neural network models, tested to forecast steel production in the world. The main goal is to find the best machine learning model that fits the steel production data in the world to make a forecast for a nine-year period. The study is important for understanding the behavior of the models and sensitivity to hyperparameters of convolutional LSTM and GRU recurrent neural networks. The results show that for long-term prediction, the GRU model is easier to train and provides better results. The article contributes to the validation of the use of other variables that are correlated with the steel production variable, thus increasing forecast accuracy.

Spatiotemporal air quality forecasting and health risk assessment over smart city of NEOM

Modeling and predicting air pollution concentrations is important to provide early warnings about harmful atmospheric substances. However, uncertainty in the dynamic process and limited information about chemical constituents and emissions sources make air-quality predictions very difficult. This study proposed a novel deep-learning method to extract high levels of abstraction in data and capture spatiotemporal features at hourly and daily time intervals in NEOM City, Saudi Arabia. The proposed method integrated a residual network (ResNet) with the convolutional long short-term memory (ConvLSTM). The ConvLSTM method was boosted by a ResNet model for deeply extracting the spatial features from meteorological and pollutant data and thereby mitigating the loss of feature information. Then, health risk assessment was put forward to evaluate PM10 and PM2.5 risk sensitivity in five districts in NEOM City. Results revealed that the proposed method with effective feature extraction could greatly optimize the accuracy of spatiotemporal air quality forecasts compared to existing state-of-the-art models. For the next hour prediction tasks, the PM10 and PM2.5 of MASE were 9.13 and 13.57, respectively. The proposed method provides an effective solution to improve the prediction of air-pollution concentrations while being portable to other regions around the world.

A Deep Intelligent Attack Detection Framework for Fog-Based IoT Systems.

Fog computing provides a multitude of end-based IoT system services. End IoT devices exchange information with fog nodes and the cloud to handle client undertakings. During the process of data collection between the layer of fog and the cloud, there are more chances of crucial attacks or assaults like DDoS and many more security attacks being compromised by IoT end devices. These network (NW) threats must be spotted early. Deep learning (DL) assumes an unmistakable part in foreseeing the end client behavior by extricating highlights and grouping the foe in the network. Yet, because of IoT devices' compelled nature in calculation and storage spaces, DL cannot be managed on those. Here, a framework for fog-based attack detection is proffered, and different attacks are prognosticated utilizing long short-term memory (LSTM). The end IoT gadget behaviour can be prognosticated by installing a trained LSTMDL model at the fog node computation module. The simulations are performed using Python by comparing LSTMDL model with deep neural multilayer perceptron (DNMLP), bidirectional LSTM (Bi-LSTM), gated recurrent units (GRU), hybrid ensemble model (HEM), and hybrid deep learning model (CNN + LSTM) comprising convolutional neural network (CNN) and LSTM on DDoS-SDN (Mendeley Dataset), NSLKDD, UNSW-NB15, and IoTID20 datasets. To evaluate the performance of the binary classifier, metrics like accuracy, precision, recall, f1-score, and ROC-AUC curves are considered on these datasets. The LSTMDL model shows outperforming nature in binary classification with 99.70%, 99.12%, 94.11%, and 99.88% performance accuracies on experimentation with respective datasets. The network simulation further shows how different DL models present fog layer communication behaviour detection time (CBDT). DNMLP detects communication behaviour (CB) faster than other models, but LSTMDL predicts assaults better.

Improving short-term bike sharing demand forecast through an irregular convolutional neural network

As an important task for the management of bike sharing systems, accurate forecast of travel demand could facilitate dispatch and relocation of bicycles to improve user satisfaction. In recent years, many deep learning algorithms have been introduced to improve bicycle usage forecast. A typical practice is to integrate convolutional (CNN) and recurrent neural network (RNN) to capture spatial–temporal dependency in historical travel demand. For typical CNN, the convolution operation is conducted through a kernel that moves across a “matrix-format” city to extract features over spatially adjacent urban areas. This practice assumes that areas close to each other could provide useful information that improves prediction accuracy. However, bicycle usage in neighboring areas might not always be similar, given spatial variations in built environment characteristics and travel behavior that affect cycling activities. Yet, areas that are far apart can be relatively more similar in temporal usage patterns. To utilize the hidden linkage among these distant urban areas, the study proposes an irregular convolutional Long-Short Term Memory model (IrConv+LSTM) to improve short-term bike sharing demand forecast. The model modifies traditional CNN with irregular convolutional architecture to leverage the hidden linkage among “semantic neighbors”. The proposed model is evaluated with a set of benchmark models in five study sites, which include one dockless bike sharing system in Singapore, and four station-based systems in Chicago, Washington, D.C., New York, and London. We find that IrConv+LSTM outperforms other benchmark models in the five cities. The model also achieves superior performance in areas with varying levels of bicycle usage and during peak periods. The findings suggest that “thinking beyond spatial neighbors” can further improve short-term travel demand prediction of urban bike sharing systems.