Search

Deep learning models have a high application value in runoff forecasting, but their prediction mechanism is difficult to interpret and their computational cost is high when dealing with complex hydrological relationships, limiting their feasibility in hydrological process mechanism analysis. To address these concerns, the paper first introduces an attention mechanism (AM) for building a long short-term memory network (LSTM) model with AM in the hidden layer (AM-LSTM). The AM-LSTM model employs attention layers to improve information extraction from hidden layers, resulting in a more accurate representation of the relationships between runoff-related elements. Furthermore, in the hidden layers of the AM-LSTM model, interpretable spatiotemporal attention units are established, which not only improves the model's prediction accuracy but also provides interpretability to the forecasting process. Furthermore, parallelization techniques are used in the paper to address the issue of model runtime cost. Simultaneously, to address the accuracy degradation caused by parallelization, the paper employs wavelet denoising (WD) techniques and builds the WD-AM-LSTM model. This accomplishment enables the runoff forecasting model to predict runoff in real time and with high accuracy. Based on validation using ten-day runoff data from the Huanren Reservoir in the Hun River's middle reaches, the results show that, with two layers and an eight-batch size, the AM-LSTM model outperforms the LSTM model in capturing spatiotemporal runoff features. During the model testing phase, the AM-LSTM model improves the MAE, RMSE, and NSE performance metrics by 8.46%, 13%, and 3.82%, respectively. The WD-AM-LSTM model effectively mitigates the noise impact caused by data parallelization under the conditions of two layers and a batch size of 512, achieving the same level of prediction performance while reducing computational cost by 92.01%. By incorporating attention mechanisms and wavelet denoising techniques, this study obtains high-speed and accurate predictions with interpretable results. It expands the deep learning models' applicability in ten-day runoff forecasting work. [ABSTRACT FROM AUTHOR]

This paper presents a comprehensive study on fall recognition and forecasting for reconfigurable stair-accessing robots by leveraging deep learning techniques. The proposed framework integrates machine learning algorithms and recurrent neural networks (RNNs), specifically Long Short-Term Memory (LSTM) and Bidirectional LSTM (BiLSTM), for fall detection of service robots on staircases. The reconfigurable stair-accessing robot sTetro serves as the platform, and the fall data required for training models are generated in a simulation environment. The two machine learning algorithms are compared and their effectiveness on the fall recognition task is reported. The results indicate that the BiLSTM model effectively classifies falls with a median categorical accuracy of 94.10% in simulation and 90.02% with limited experiments. Additionally, the BiLSTM model can be used for forecasting, which is practically valuable for making decisions well before the onset of a free fall. This study contributes insights into the design and implementation of fall detection systems for service robots used to navigate staircases through deep learning approaches. Our experimental and simulation data, along with the simulation steps, are available for reference and analysis via the shared link. [ABSTRACT FROM AUTHOR]

Photovoltaic (PV) power generation forecasting is an important research topic, aiming to mitigate variability caused by weather conditions and improve power generation planning. Climate factors, including solar irradiance, temperature, and cloud cover, influence the energy conversion achieved by PV systems. Long-term weather forecasting improves PV power generation planning, while short-term forecasting enhances control methods, such as managing ramp rates. The stochastic nature of weather variables poses a challenge for linear regression methods. Consequently, advanced, state-of-the-art machine learning (ML) approaches capable of handling non-linear data, such as long short-term memory (LSTM), have emerged. This paper introduces the implementation of a multivariate machine learning model to forecast PV power generation, considering multiple weather variables. A deep learning solution was implemented to analyze weather variables in a short time horizon. Utilizing a hidden Markov model for data preprocessing, an LSTM model was trained using the Alice Spring dataset provided by DKA Solar Center. The proposed workflow demonstrated superior performance compared to the results obtained by state-of-the-art methods, including support vector machine, radiation classification coordinate with LSTM (RCC-LSTM), and ESNCNN specifically concerning the proposed multi-input single-output LSTM model. This improvement is attributed to incorporating input features such as active power, temperature, humidity, horizontal and diffuse irradiance, and wind direction, with active power serving as the output variable. The proposed workflow achieved a mean square error (MSE) of 2.17 × 10 − 7 , a root mean square error (RMSE) of 4.65 × 10 − 4 , and a mean absolute error (MAE) of 4.04 × 10 − 4 . [ABSTRACT FROM AUTHOR]

Tennis has gained global popularity, prompting a surge in interest towards 3D video-based tennis motion recognition. Early action recognition, which predates activity completion, is a critical classification task to preempt adverse outcomes. Prior research emphasizes effective feature extraction and modeling for swift, accurate classification, despite limited data availability. To establish a robust foundation, this study introduces an anticipatory action prediction module preceding the recognition component. The module forecasts subsequent motions based on observed ones, using an LSTM-GAN structure to mitigate motion blurring and generate predictions. This paper presents an innovative framework that leverages deep learning, particularly dilated neural networks, for real-time spatio-temporal tennis analysis on standard hardware, aiming to enhance player performance insights and action prediction through TensorFlow. The dilated RNN and CNN are integrated into the recognition module for comprehensive spatiotemporal feature modeling. To foster synergy between the prediction and recognition modules, a hard class mining mechanism is devised to enhance the learning capabilities of challenging class samples. As a result, the LSTM architecture combined with GAN provides an excellent 92.1 Precision, 91.2 Recall, 94.5 F-1 score and 95.0 Accuracy in action recognition and prediction of tennis sport, which is significantly higher than classical models i.e. GAN, Conv3DJ, Co-occurrence LSTM, and GAN + L1 + Mining. [ABSTRACT FROM AUTHOR]

Day by day pollution in cities is increasing due to urbanization. One of the biggest challenges posed by the rapid migration of inhabitants into cities is increased air pollution. Sustainable Development Goal 11 indicates that 99 percent of the world's urban population breathes polluted air. In such a trend of urbanization, predicting the concentrations of pollutants in advance is very important. Predictions of pollutants would help city administrations to take timely measures for ensuring Sustainable Development Goal 11. In data engineering, imputation and the removal of outliers are very important steps prior to forecasting the concentration of air pollutants. For pollution and meteorological data, missing values and outliers are critical problems that need to be addressed. This paper proposes a novel method called multiple iterative imputation using autoencoder-based long short-term memory (MIA-LSTM) which uses iterative imputation using an extra tree regressor as an estimator for the missing values in multivariate data followed by an LSTM autoencoder for the detection and removal of outliers present in the dataset. The preprocessed data were given to a multivariate LSTM for forecasting PM2.5 concentration. This paper also presents the effect of removing outliers and missing values from the dataset as well as the effect of imputing missing values in the process of forecasting the concentrations of air pollutants. The proposed method provides better results for forecasting with a root mean square error (RMSE) value of 9.8883. The obtained results were compared with the traditional gated recurrent unit (GRU), 1D convolutional neural network (CNN), and long short-term memory (LSTM) approaches for a dataset of the Aotizhonhxin area of Beijing in China. Similar results were observed for another two locations in China and one location in India. The results obtained show that imputation and outlier/anomaly removal improve the accuracy of air pollution forecasting. [ABSTRACT FROM AUTHOR]

• The paper fills the gap in the literature by reviewing and analyzing the current studies that apply deep learning for Covid-19 forecasting. • The initial search identified 152 studies of which 53 passed the quality control. • The existing literature is categorized using a model-based taxonomy. • The description of the models along with their performance evaluation is presented. • Recommendations for future improvements are provided. The Covid-19 pandemic has galvanized scientists to apply machine learning methods to help combat the crisis. Despite the significant amount of research there exists no comprehensive survey devoted specifically to examining deep learning methods for Covid-19 forecasting. In this paper, we fill the gap in the literature by reviewing and analyzing the current studies that use deep learning for Covid-19 forecasting. In our review, all published papers and preprints, discoverable through Google Scholar, for the period from Apr 1, 2020 to Feb 20, 2022 which describe deep learning approaches to forecasting Covid-19 were considered. Our search identified 152 studies, of which 53 passed the initial quality screening and were included in our survey. We propose a model-based taxonomy to categorize the literature. We describe each model and highlight its performance. Finally, the deficiencies of the existing approaches are identified and the necessary improvements for future research are elucidated. The study provides a gateway for researchers who are interested in forecasting Covid-19 using deep learning. [ABSTRACT FROM AUTHOR]

GNSS time series prediction plays a significant role in monitoring crustal plate motion, landslide detection, and the maintenance of the global coordinate framework. Long short-term memory (LSTM) is a deep learning model that has been widely applied in the field of high-precision time series prediction and is often combined with Variational Mode Decomposition (VMD) to form the VMD-LSTM hybrid model. To further improve the prediction accuracy of the VMD-LSTM model, this paper proposes a dual variational modal decomposition long short-term memory (DVMD-LSTM) model to effectively handle noise in GNSS time series prediction. This model extracts fluctuation features from the residual terms obtained after VMD decomposition to reduce the prediction errors associated with residual terms in the VMD-LSTM model. Daily E, N, and U coordinate data recorded at multiple GNSS stations between 2000 and 2022 were used to validate the performance of the proposed DVMD-LSTM model. The experimental results demonstrate that, compared to the VMD-LSTM model, the DVMD-LSTM model achieves significant improvements in prediction performance across all measurement stations. The average RMSE is reduced by 9.86% and the average MAE is reduced by 9.44%; moreover, the average R2 increased by 17.97%. Furthermore, the average accuracy of the optimal noise model for the predicted results is improved by 36.50%, and the average velocity accuracy of the predicted results is enhanced by 33.02%. These findings collectively attest to the superior predictive capabilities of the DVMD-LSTM model, thereby demonstrating the reliability of the predicted results. [ABSTRACT FROM AUTHOR]

[Display omitted] • Using LSTM and GSA-LSTM methods to forecast PV power in different horizons. • Proposes more precise hyperparameters for LSTM to improve performance prediction. • Couples LSTM with GSA to optimize hyperparameters, enhances prediction accuracy. • GSA-LSTM is superior by improving MSE at peak points compared to LSTM. This research paper investigates the use of Long Short-Term Memory (LSTM) and Grid Search Algorithm (GSA)-LSTM methods to forecast PV power output in different horizons. The study proposes precise hyperparameters for the LSTM network for improved performance prediction. Initially, LSTM was evaluated solely, and three different scenarios were investigated to obtain the best LSTM network hyperparameter settings for various horizons using a sensitivity study. Afterwards, LSTM was coupled with GSA to optimize LSTM network hyperparameters, which enhanced prediction accuracy and minimized error. The proposed methodology identified the most effective values for each hyperparameter based on different forecasting horizons. Additionally, Spearman Correlation Coefficient (SCC) and Pearson Correlation Coefficient (PCC) were deployed to determine the relationship between input data and target as well as identifying optimum lag value which has substantial effect on LSTM network. Optimal lag values of 150 and 280 were determined following the strong correlation between the data and target. LSTM and GSA-LSTM methods were compared with each other at various horizons and it was demonstrated that GSA-LSTM is superior by improving MSE at peak points by 10%, 30%, and 34% for 12 hours, 3 days, and 2 weeks horizons respectively. Comparing the proposed methodology with other studies in the literature revealed that the current study is capable of predicting PV power output for the 1 hour ahead horizon with significantly higher accuracy and with maximum improvement of about 28%. [ABSTRACT FROM AUTHOR]

The Remaining Useful Life of a machine is very useful statistical information for the operator and manufacturer. It provides a very clear perspective to the user how long the machine can be operated and if any faults are detected how can they be prevented and ultimately increase the Remaining Useful Life. If the operators are aware of the forthcoming issues of the machine the downtime caused in the inspection, part delivery and eventually replacing parts is significantly reduced. The paper presents a study on the remaining useful life of machinery as it is an emerging technique, starting from the year 2010 to the current year of 2021. The bibliometric research is carried out using the Scopus database and software such as the VOS Viewer and GPS Visualizer. The study tells that the maximum publications are from articles and conference papers affiliated to remaining useful life, Chinese lead the publication followed by the United States and then France. Remaining useful lives is the most used keyword followed by forecasting and the least contributing keywords are also found. [ABSTRACT FROM AUTHOR]

Renewable energies, such as solar and wind power, have become promising sources of energy to address the increase in greenhouse gases caused by the use of fossil fuels and to resolve the current energy crisis. Integrating wind energy into a large-scale electric grid presents a significant challenge due to the high intermittency and nonlinear behavior of wind power. Accurate wind power forecasting is essential for safe and efficient integration into the grid system. Many prediction models have been developed to predict the uncertain and nonlinear time series of wind power, but most neglect the use of Bayesian optimization to optimize the hyperparameters while training deep learning algorithms. The efficiency of grid search strategies decreases as the number of hyperparameters increases, and computation time complexity becomes an issue. This paper presents a robust and optimized long-short term memory network for forecasting wind power generation in the day ahead in the context of Ethiopia's renewable energy sector. The proposal uses Bayesian optimization to find the best hyperparameter combination in a reasonable computation time. The results indicate that tuning hyperparameters using this metaheuristic prior to building deep learning models significantly improves the predictive performances of the models. The proposed models were evaluated using MAE, RMSE, and MAPE metrics, and outperformed both the baseline models and the optimized gated recurrent unit architecture. [ABSTRACT FROM AUTHOR]

Storage workload prediction is a critical step for fine-grained load balancing and job scheduling in realtime and adaptive cluster systems. However, how to perform workload time series prediction based on a deep learning method has not yet been thoroughly studied. In this paper, we propose a storage workload prediction method called CrystalLP based on deep learning. CrystalLP includes workload collecting, data preprocessing, time series prediction, and data post-processing phase. The time series prediction phase is based on a long short-term memory network (LSTM). Furthermore, to improve the efficiency of LSTM, we study the sensitivity of the hyperparameters in LSTM. Extensive experimental results show that CrystalLP can obtain performance improvement compared with three classic time series prediction algorithms. [ABSTRACT FROM AUTHOR]

Drastic hog price fluctuations have a great impact on the welfare of hog farmers, people's living standards, and the macroeconomy. To stabilise the hog price, hog price forecasting has become an increasingly hot issue in the research literature. Existing papers have neglected the benefits of decomposition and instead directly utilise models to predict hog prices by capturing raw data. Motivated by this issue, the authors introduce a new robust forecasting approach for hog prices that combines ensemble empirical mode decomposition (EEMD) and multilong short-term memory neural networks (Multi-LSTMs). First, EEMD decomposes the volatile raw sequence into several smoother subsequences. Second, the decomposed subsequences are predicted separately using a parallel structure model consisting of several LSTMs. Finally, the fuse function combines all the subresults to yield the final result. The empirical results suggest that the proposed method only has minor errors and proves the effectiveness and reliability in experiments on real datasets (2.55207, 4.816, and 0.332 on MAE, MAPE and RMSLE, respectively). Reliable forecasting of hog prices is beneficial to farmers and people to allow optimisation of their production and booking rates and to moderate the adverse effects of potential shocks. [ABSTRACT FROM AUTHOR]

This paper investigates a deep learning‐based wind‐forecasting model to establish an accurate forecasting model which can support the increasing growth of wind power generation. The wind forecasting means wind speed and direction forecasting at the same time. Proposed forecasting model consists of three‐dimensional convolutional neural network and deep convolutional long short‐term memory (3DCNN‐DConvLSTM), and forecasts the wind vector which expressed as time‐sequential images. DConvLSTM model learns spatiotemporal features from time‐series image data that represent a spatial and temporal change of wind speed and direction. The forecasting model combined of 3DCNN and DConvLSTM is effective to decrease training time, and forecasting error in comparison to the DConvLSTM model. Input of the forecasting model is wind speed and direction that is expressed as an image on 2D coordinate and uses the measured data by the Automated Meteorological Data Acquisition System (AMeDAS), Japan. Forecasting accuracy with one‐hour ahead and its usefulness of the proposed forecasting model is evaluated with simulation results for four seasons that is typical of Japan climate, and demonstrated by comparison with fully connected‐LSTM (FC‐LSTM), encoder‐decoder based 3DCNN (ED‐3DCNN), DConvLSTM, and persistent models. © 2022 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC. [ABSTRACT FROM AUTHOR]

Forecasting return and profit is a primary challenge for financial practitioners and an even more critical issue when it comes to forecasting energy market returns. This research attempts to propose an effective method to predict the Brent Crude Oil return, which results in remarkable performance compared with the well-known models in the return prediction. The proposed hybrid model is based on long short-term memory (LSTM) and convolutional neural network (CNN) networks where the autoregressive integrated moving average (ARIMA) and generalized autoregressive conditional heteroscedasticity (GARCH) outputs are used as features, along with return lags, price, and macroeconomic variables to train the models, resulting in significant improvement in the model's performance. According to the obtained results, our proposed model performs better than other models, including artificial neural network (ANN), principal component analysis (PCA)-ANN, LSTM, and CNN. We show the efficiency of our proposed model by testing it with a simple trading strategy, indicating that the cumulative profit obtained from trading with the prediction results of the proposed 2D CNN-LSTM model is higher than those of the other models presented in this research. In the second part of this study, we consider the spread of COVID-19 and its impact on the financial markets to present a precise LSTM model that can reflect the impact of this disease on the Brent Crude Oil return. This paper uses the significance test and correlation measures to show the similarity between the series of Brent Crude Oil during the SARS and the COVID-19 pandemics, after which the data during the SARS period are used along with the data during COVID-19 to train the LSTM. The results demonstrate that the proposed LSTM model, tuned by the SARS data, can better predict the Brent Crude Oil return during the COVID-19 pandemic. [ABSTRACT FROM AUTHOR]

Predicting volatility is a critical activity for taking risk- adjusted decisions in asset trading and allocation. In order to provide effective decision-making support, in this paper we investigate the profitability of a deep Long Short-Term Memory (LSTM) Neural Network for forecasting daily stock market volatility using a panel of 28 assets representative of the Dow Jones Industrial Average index combined with the market factor proxied by the SPY and, separately, a panel of 92 assets belonging to the NASDAQ 100 index. The Dow Jones plus SPY data are from January 2002 to August 2008, while the NASDAQ 100 is from December 2012 to November 2017. If, on the one hand, we expect that this evolutionary behavior can be effectively captured adaptively through the use of Artificial Intelligence (AI) flexible methods, on the other, in this setting, standard parametric approaches could fail to provide optimal predictions. We compared the volatility forecasts generated by the LSTM approach to those obtained through use of widely recognized benchmarks models in this field, in particular, univariate parametric models such as the Realized Generalized Autoregressive Conditionally Heteroskedastic (R-GARCH) and the Glosten–Jagannathan–Runkle Multiplicative Error Models (GJR-MEM). The results demonstrate the superiority of the LSTM over the widely popular R-GARCH and GJR-MEM univariate parametric methods, when forecasting in condition of high volatility, while still producing comparable predictions for more tranquil periods. [ABSTRACT FROM AUTHOR]

In predicting energy consumption, classic econometric and statistical models are used to forecast energy consumption. These models may have limitations in an increasingly fast-changing energy market, which requires big data analysis of energy consumption patterns and relevant variables using complex mathematical tools. In current literature, there are minimal comparison studies reviewing machine learning algorithms to predict energy consumption in Korea. To bridge this gap, this paper compared three different machine learning algorithms, namely the Random Forest (RF) model, XGBoost (XGB) model, and Long Short-Term Memory (LSTM) model. These algorithms were applied in Period 1 (prior to the onset of the COVID-19 pandemic) and Period 2 (after the onset of the COVID-19 pandemic). Period 1 was characterized by an upward trend in energy consumption, while Period 2 showed a reduction in energy consumption. LSTM performed best in its prediction power specifically in Period 1, and RF outperformed the other models in Period 2. Findings, therefore, suggested the applicability of machine learning to forecast energy consumption and also demonstrated that traditional econometric approaches may outperform machine learning when there is less unknown irregularity in the time series, but machine learning can work better with unexpected irregular time series data. [ABSTRACT FROM AUTHOR]

The deep learning (DL) approaches in smart grid (SG) describes the possibility of shifting the energy industry into a modern era of reliable and sustainable energy networks. This paper proposes a time-series clustering framework with multi-step time-series sequence to sequence (Seq2Seq) long short-term memory (LSTM) load forecasting strategy for households. Specifically, we investigate a clustering-based Seq2Seq LSTM electricity load forecasting model to undertake an energy load forecasting problem, where information input to the model contains individual appliances and aggregate energy as historical data of households. The original dataset is preprocessed, and forwarded to a multi-step time-series learning model which reduces the training time and guarantees convergence for energy forecasting. Furthermore, simulation results show the accuracy performance of the proposed model by validation and testing cluster data, which shows a promising potential of the proposed predictive model. [ABSTRACT FROM AUTHOR]

Drastic hog price fluctuations have a great impact on the welfare of hog farmers, people's living standards, and the macroeconomy. To stabilise the hog price, hog price forecasting has become an increasingly hot issue in the research literature. Existing papers have neglected the benefits of decomposition and instead directly utilise models to predict hog prices by capturing raw data. Motivated by this issue, the authors introduce a new robust forecasting approach for hog prices that combines ensemble empirical mode decomposition (EEMD) and multilong short-term memory neural networks (Multi-LSTMs). First, EEMD decomposes the volatile raw sequence into several smoother subsequences. Second, the decomposed subsequences are predicted separately using a parallel structure model consisting of several LSTMs. Finally, the fuse function combines all the subresults to yield the final result. The empirical results suggest that the proposed method only has minor errors and proves the effectiveness and reliability in experiments on real datasets (2.55207, 4.816, and 0.332 on MAE, MAPE and RMSLE, respectively). Reliable forecasting of hog prices is beneficial to farmers and people to allow optimisation of their production and booking rates and to moderate the adverse effects of potential shocks. [ABSTRACT FROM AUTHOR]

• We discuss the techniques for finding anomalies in multi-dimensional time series. • Statistical strategy to remove the constraints of high dimensionality in data and choose more significant characteristics. • In high-dimensional datasets, anomaly identification has improved using deep learning techniques. • The original meanings of attributes are restored while choosing those that have statistical significance. Recent advances in time series forecasting and anomaly detection have been attributed to the growing popularity of deep learning approaches. Traditional methods, such as rule-based systems and statistical techniques, have limitations when applied to complex and dynamic real-world data. This study investigates using various deep learning models for anomaly detection, recognising aberrant patterns in data, and time series forecasting. The performance of the proposed models is evaluated on benchmarks like the Numenta Anomaly Benchmark (NAB) corpus and credit card fraud detection, showing their ability to detect aberrant patterns in various scenarios. Preprocessing strategies, such as normalisation and feature scaling, play a significant role in both time series forecasting and anomaly detection. In addition, the paper proposes a statistical method for selecting different or more important features from a dataset to overcome the limitations of high-dimensional sequencing data. In many ways, the suggested feature selection technique outperforms previous solutions. It keeps the original meanings of the attributes while selecting those with statistical relevance. Furthermore, it is computationally efficient and successfully solves the problem of excessive dimensions. Overall, deep learning approaches for time series forecasting and anomaly detection are promising in banking, healthcare, and manufacturing industries. [ABSTRACT FROM AUTHOR]

This paper aims to incorporate the throttle handle level prediction in high speed train(HST) operation prediction problem to enable the prediction of HST drivers' activities, in which the key instructions available to HST driver are difficult to determine. Specifically, we consider an end-edge-cloud orchestration system to capture the real-time responses for driver state changes. By adding edge computing nodes, the real-time performance of data collection, transmission, and processing is improved. Our ultimate goal is to guide and regulate train drivers' activities in the same way, regardless of uncertain factors affecting HST dynamic or kinematic performance. We formulate the problem as a physical-based and data-driven deep learning-aided prediction model and solve it using a novel long short-term memory (LSTM) deep neural network which combines: (i) an off-line approximate training model to learn the time series data in the cloud layer, and (ii) an online prediction process to determine driving strategies in the real-time windows, more in general expressed as driving skill level constraints. To evaluate the performance of our approach, some case studies using the real-world railway infrastructure and HST data have been conducted. The results show that the proposed models produce higher prediction accuracy for both speed and throttle handle level prediction tasks. Compared to the conventional HST operation prediction problem, which considers speed sequences only without throttle handle level consideration, this study finds that jointly modeling speed and throttle handle level actually improves the next operation prediction performance itself, potentially because throttle handle level observations capture the information on HST control dynamics, which may affect operators' driving choices. • Structure. Edge computing and prediction scheme are innovatively combined with HST data collection to meet the delay-sensitive and sequential-aware operation requirements, and a novel HST operation structure for end-edge cloud orchestration system is built. • Modeling. A combination of the physical-based and the data-driven deep learning model is formulated for HST operation prediction under uncertainty in throttle handle level and speed, which is a new model in the literature. As discussed, the existing research has only considered speed dynamics in real-time DAS limitedly. • Practice. From the field collected data in Beijingxi–Zhengzhoudong HSR (as shown in Figure 5), we test the effectiveness of our proposed model with seven cases. Since the real-time driving advice is calculated in a few milliseconds, our two-phase solution method can be practically relevant when designing DAS or ATO systems. More importantly, most time-consuming calculations are executed off-line in the cloud layer and the amount of on-line calculations in edge layer is limited to looking up a table. [ABSTRACT FROM AUTHOR]

Transmission lines' condition monitoring is an important part of smart grid construction. To ensure fast and efficient transmission of data, many mash-based wireless networks devices are adopted to collect status information. Since these nodes are exposed to the natural environment, vulnerable to damage, so it is very necessary to be predicting nodes' fault. However, these mesh nodes are affected by a variety of complex and time-series factors, and traditional models are difficult to achieve effective failure prediction. To solve this problem, this paper proposes a self-adapting multi-LSTM ensemble regression model for transmission line network's wireless mesh node failure prediction (MLSTM-FP), through establishes the corresponding relationship between similar time factors and LSTMs, the proposed model can realize multi time series data self-adapting and accurate failure prediction of transmission line network's wireless mesh nodes, The experimental results show that the proposed method has a good prediction ability than traditional methods. [ABSTRACT FROM AUTHOR]

Owing to integrating photovoltaic solar systems into power networks, accurate prediction of solar irradiance plays an increasingly significant role in electric energy planning and management. However, the existing hybrid models ignore the influence of other factors except for the irradiance time series and adopt a single branch independent network structure, which may lead to decrease prediction accuracy. In this paper, a novel multivariate hybrid deep neural model named WPD–CNN–LSTM-MLP for 1-h-ahead solar irradiance forecasting is proposed. The novel WPD–CNN–LSTM-MLP model is based on a sophisticated multi-branch hybrid structure with multi-variable inputs, which the multi-branch hybrid structure combines wavelet packet decomposition (WPD), convolutional neural network (CNN), long short-term memory (LSTM) networks, and multi-layer perceptron network (MLP), and the multi-variable inputs include hourly solar irradiance and three climate variables, namely: temperature, relative humidity, and wind speed and their combination. The new model extracts the inherent characteristics of multi-layer inputs sufficiently, overcomes the shortcomings of traditional models, and achieves more accurate forecasting results. The performance of the model is verified by actual data from Denver, Clark, and Folsom, the United States. Comparative studies of traditional individual back propagation neural network, support vector machine, recurrent neural network, LSTM, the climatology–persistence reference forecasts method and the proposed LSTM-MLP model, CNN-LSTM-MLP model, and WPD–CNN–LSTM model reveal that the proposed WPD–CNN–LSTM-MLP deep learning model has better prediction accuracy in hourly irradiance forecasting. • Hybrid deep learning model is proposed for hourly solar irradiance forecasting. • A multi-branch hybrid structure with multivariable inputs is developed. • WPD, CNN, LSTM, and MLP are combined. • Irradiance prediction using both frequency and spatial features for the first time. • The multi-branch structure with multivariable inputs shows excellent performance. [ABSTRACT FROM AUTHOR]

Electricity is an essential resource for human production and survival. Accurately predicting electrical load consumption can help power supply companies make informed decisions, such as peak load shifting, to maintain a reliable power supply and reduce CO2 emissions. However, forecasting electricity consumption is challenging due to the nonlinear and nonstationary time series data that is correlated with climate change. To address this challenge, this paper proposes an electricity forecasting method based on empirical mode decomposition (EMD) and bidirectional LSTM. EMD is a solid and robust instrument for time–frequency analysis and signal preprocessing, which separates the time series into components at different resolutions. The proposed model predicts the future 24 h with a resolution of 15 min by creating many stationary component sequences from the original stochastic electricity usage time series data (IMFs). To predict each Intrinsic Mode Function, a hybrid model BI-LSTM is employed. The results of each component's forecast are then merged to give the overall forecast. Two comparative studies are conducted to justify the choice of the signal processing method and the prediction algorithm. The proposed model demonstrates a minimal MAPE of 0.28% and a better R 2 close to 1 of 0.84 compared to other papers. [ABSTRACT FROM AUTHOR]

Multivariate time series (MTS) forecasting is a research field that is gaining more and more importance as time series data generators proliferate in the growing era of Internet of Things. Deep learning architecture for MTS data has been and still a very active research area as there is no comprehensive comparative study of the different architectures, let alone a perfect architecture that can solve all types of time series modeling problem. In this paper, we start by highlighting time series analysis requirements, to propose a new model for conditional multivariate time series forecasting based on the lately introduced WaveNet architecture. Our model is composed of stacked residual causal dilated convolutions that provide large scope in time series history and foster learning of long-term dependencies. Parameterized skip connections allow catching early trends and properties while conditioning enables modeling the associations that exist between multivariate data. We used group normalization for its stability and independence from the batch size. We present a deep comparison of the structure, flexibility, flow control, memory consumption, robustness and stability capacities of our new model against those of recurrent neural networks state-of-the-art approaches—long-short-term-memory (LSTM) and gated recurrent unit (GRU). To assess the performances of our model, we conduct extensive experimental testing on the task of urban air quality prediction in Marrakesh city using six real-world multi-sensor and multivariate time series datasets. We compare the results of our model against the results of LSTM and GRU. Our experiments revealed that our proposed WaveNet-temporal-CNN outperform recurrent models ability to learn long-term dependencies in a time-efficient way. [ABSTRACT FROM AUTHOR]

Short-term load forecasting (STLF) is a critical component of smart grid operations, yet it is a challenging task due to the high uncertainty of electrical loads. This paper proposes a novel STLF model by combining the fuzzy c-means (FCM) clustering and an improved long short-term memory (LSTM) neural network. The load profiles of two consecutive days are extracted as a single sample and their dimension is reduced by principal component analysis (PCA). The FCM clustering algorithm is then used to group the load profiles into similar patterns from a historical load data set. For each pattern, an LSTM-based forecaster is constructed and optimized using the load profiles that belong to it. The periodicity of the load profiles at the same time of two days is taken into account when designing the forecaster, resulting in a new LSTM model. The experimental results on two commonly used electrical load data sets demonstrate superior effectiveness and performance compared to other models in terms of the MAPE metric. [Display omitted] • A novel hybrid short-term load forecasting model is proposed by combining clustering and deep learning. • A new long short-term memory neural network is designed considering the periodicity of load profiles. • The proposed STLF model achieves much lower MAPE values than others. [ABSTRACT FROM AUTHOR]

Deep learning has made significant advances in methodologies and practical applications in recent years. However, there is a lack of understanding on how the long short-term memory (LSTM) networks perform in river flow prediction. This paper assesses the performance of LSTM networks to understand the impact of network structures and parameters on river flow predictions. Two river basins with different characteristics, i.e., Hun river and Upper Yangtze river basins, are used as case studies for the 10-day average flow predictions and the daily flow predictions, respectively. The use of the fully connected layer with the activation function before the LSTM cell layer can substantially reduce learning efficiency. On the contrary, non-linear transformation following the LSTM cells is required to improve learning efficiency due to the different magnitudes of precipitation and flow. The batch size and the number of LSTM cells are sensitive parameters and should be carefully tuned to achieve a balance between learning efficiency and stability. Compared with several hydrological models, the LSTM network achieves good performance in terms of three evaluation criteria, i.e., coefficient of determination, Nash-Sutcliffe Efficiency and relative error, which demonstrates its powerful capacity in learning non-linear and complex processes in hydrological modelling. [ABSTRACT FROM AUTHOR]

Anomaly detection in power consumption data can be very useful to building managers. It allows them to detect unexpected power consumption values, identify unusual behaviors, and foresee uncommon events. This paper proposes a novel unsupervised approach to detect anomalies in power consumption data. We combine the clustering-based methods with the prediction-based ones to learn typical behavior scenarios and to predict the power consumption of the next hour. These scenarios are explored by applying the K-means algorithm on 24 different K-means groups representing the 24 h of the day. This is based on the assumption that identical daily consumption behavior can appear repeatedly due to users' living habits. In order to detect the anomaly 1 h before its occurrence, a Long Short-Term Memory (LSTM) has been trained to predict the next power consumption value. This predicted value with some earlier data values are concatenated into a vector then compared with the learned typical scenarios. We used Auto-Encoders to detect anomalous days in general and this novel method to specify at what time the anomaly has occurred. Our approach not only detectss anomalies in off-line mode but also allows real-time detection on live data streams. [ABSTRACT FROM AUTHOR]

Stock market timing is regarded as a challenging task of financial prediction. An accurate prediction of stock trend can yield great profits for investors. At present, recurrent neural networks (RNNs) have a good performance in stock market forecasting. However, there has been a relative lack of research in the stock market timing using RNNs. In this paper, a novel model named hybrid RNN model is proposed for stock market timing by incorporating multi-layer long short-term memory, multi-layer gated recurrent unit and one-layer ReLU layer. Moreover, based on five popular benchmark datasets from UCI Machine Learning Repository and six daily securities from Shanghai Stock Exchange, comparisons with 12 state-of-the-art models are conducted to verify the superiority of the proposed hybrid RNN model in terms of nine technical indicators. The findings from the experiment demonstrate that: (1) as opposed to 12 models, the average accuracy, MSE and AUC of hybrid RNN model (0.7406, 0.2592, 0.7368) are significantly better than other comparison models, and (2) the proposed hybrid RNN classification procedure can be considered as a feasible and effective tool for stock market timing. [ABSTRACT FROM AUTHOR]

• Forecasting models for solid waste generation at a household level were developed. • A multi-site Long Short-Term Memory neural network approach is proposed. • Models were tested on real-life data of weekly waste generation rates from Denmark. • The multi-site LSTM model outperforms traditional models by 85% on average. Forecasting household waste generation using traditional methods is particularly challenging due to its high variability and uncertainty. Unlike studies that forecast waste generation at municipal or country levels, household data can present rapid short-term variations and highly non-linear dynamics. The aim of this paper is to investigate the advantages of using a state-of-the-art deep learning approach compared to traditional forecasting methods. We apply a multi-site Long Short-Term Memory (LSTM) Neural Network, to forecast waste generation rates from households using a long-term data base. The model is applied to historical data of weekly waste weights from households in the municipality of Herning, Denmark, in the period between 2011 and 2018. Results show that using a multi-site approach, instead of an individual fit for each household, can improve forecasting performance of the LSTM model by 28% on average, and that the LSTM approaches can effectively improve the results by 85% on average compared with traditional methods such as ARIMA. [ABSTRACT FROM AUTHOR]

Due to the inaccuracy and significant disturbance of the complex and harsh environment in real industrial processes, the traditional sensor devices cannot meet the high-performance requirement of measuring key quality variables. However, in practical industrial thickener cone systems, the underflow concentration is hard to measure and has a high cost and significant time delay. Furthermore, the higher encoder representation often causes information loss from the process variables. This paper presents a novel efficient dual long short-time memory (LSTM) method for concentration prediction in the deep cone thickener system. To this end, dual feedforward and inverse bidirectional long-short time memory are proposed for feature learning and long temporal prediction. The proposed framework introduces an averaging moving filtering to pass through features, therefore the performance of dual LSTM is increased by a large margin. In addition, the feedforward and reverse bidirectional LSTM are employed to learn the robust information without loss. At last, experimental verification of the performance of an industrial deep cone thickener demonstrates the proposed dual LSTM method outperforms other state-of-the-art methods. • Proposing a new prediction model architecture, DualLSTM. • Utilizing the feedforward and reverse BiLSTM as the basic unit, DualLSTM's performance improved. • Designing a novel back-propagation through time (BPTT) algorithm for the DualLSTM. • Employing the DualLSTM to tackle the underflow concentration prediction in industrial CTS. [ABSTRACT FROM AUTHOR]

• The Transductive Long Short Term Memory model is proposed. • The proposed result indicates that the T -LSTM is better than previous studies. • The T -LSTM method produces 98.2% accuracy in forecasting the weather. • This proposed hybrid T -LSTM method provides a robust solution. Forecasting climate and the development of the environment have been essential in recent days since there has been a drastic change in nature. Weather forecasting plays a significant role in decision-making in traffic management, tourism planning, crop cultivation in agriculture, and warning the people nearby the seaside about the climate situation. It is used to reduce accidents and congestion, mainly based on climate conditions such as rainfall, air condition, and other environmental factors. Accurate weather prediction models are required by meteorological scientists. The previous studies have shown complexity in terms of model building, and computation, and based on theory-driven and rely on time and space. This drawback can be easily solved using the machine learning technique with the time series data. This paper proposes the state-of-art deep learning model Long Short-Term Memory (LSTM) and the Transductive Long Short-Term Memory (T -LSTM) model. The model is evaluated using the evaluation metrics root mean squared error, loss, and mean absolute error. The experiments are carried out on HHWD and Jena Climate datasets. The dataset comprises 14 weather forecasting features including humidity, temperature, etc. The T -LSTM method performs better than other methodologies, producing 98.2% accuracy in forecasting the weather. This proposed hybrid T -LSTM method provides a robust solution for the hydrological variables. [ABSTRACT FROM AUTHOR]

Long-short term memory (LSTM) network is one of the state-of-the-art models to forecast the movement of financial time series (FTS). However, existing LSTM networks do not perform well in the long-term forecasting FTS with sharp change points, which significantly influences the accumulated returns. This paper proposes a novel long-term forecasting method of FTS movement based on a modified adaptive LSTM model. The adaptive network mainly consists of two LSTM layers followed by a pair of batch normalization (BN) layers, a dropout layer and a binary classifier. In order to capture the important profit points, we propose to use an adaptive cross-entropy loss function that enhances the prediction capacity on the sharp changes and deemphasizes the slight oscillations. Then, we perform the forecasting on multiple independent networks and vote on their output data to obtain stable forecasting result. Considering the temporal correlation of FTS, an inherited training strategy is introduced to accelerate the retraining procedure when performing the long-term forecasting task. The proposed methods are assessed and verified by the numerical experiments on the stock index datasets, including "Standard's & Poor's 500 Index", "China Securities Index 300" and "Shanghai Stock Exchange 180". A substantial improvement of forecasting performance is proved. Moreover, the proposed hybrid forecasting framework can be generalized to different FTS datasets and deep learning models. • LSTM network is a useful model to forecast the movement of financial time series. • The proposed forecasting framework can bring more accumulated returns. • The adaptive loss function enhances the forecasting capacity on the sharp changes. • A stable result is obtained using vote operation on multiple independent results. • An inherited training accelerates the fine-turn procedure in a long-term prediction. [ABSTRACT FROM AUTHOR]

• Integrating financial conditions with LSTM improves CDS forecasting performance. • Merton-LSTM outperforms GRU, SVM, MLP and stochastic series model in forecasting CDS. • The Merton-LSTM trading strategy eclipses other tested trading strategies. • The superiority of the Merton-LSTM model proved in long-term prediction. Using macroeconomic and financial conditions to forecast credit default swap (CDS) spreads is a challenging task. In this paper, we propose the Merton-LSTM model, a modified LSTM model formed by integrating with the Merton determinants model, to forecast the CDS indices. We provide the rigorous math behind the Merton-LSTM model, which demonstrates that by leveraging the nonlinear learning ability of LSTM with increased model capacity, the Merton-LSTM model is expected to learn the inherent association between the Merton determinants and CDS spreads. Further, the Merton-LSTM model is compared with the machine learning models LSTM, gated recurrent unit (GRU), multilayer perceptron network (MLP), support vector machine (SVM) and a typical stochastic series model in forecasting the two most liquid five-year CDS indices, North America High Yield index (CDX.NA.HY) and North America Investment Grade index (CDX.NA.IG) through the root mean squared error (RMSE) and the Diebold-Mariano test. The comparison results show that the RMSEs of the Merton-LSTM model are the lowest (6.2570–27.2000 for CDX.NA.HY and 1.3168–6.4772 for CDX.NA.IG) compared to other competitive models. The superiority of the Merton-LSTM model in forecasting performance is highlighted in long-term prediction even with a forecasting horizon extended to 28 days. Simulated trading with different thresholds and horizons is conducted in this study. We find that the Merton-LSTM trading strategy yields the highest annualized Sharpe ratios and lowest maximum losses at most thresholds and horizons, highlighting the economic significance of the proposed model. [ABSTRACT FROM AUTHOR]

• Integrating financial conditions with LSTM improves CDS forecasting performance. • Merton-LSTM outperforms GRU, SVM, MLP and stochastic series model in forecasting CDS. • The Merton-LSTM trading strategy eclipses other tested trading strategies. • The superiority of the Merton-LSTM model proved in long-term prediction. Using macroeconomic and financial conditions to forecast credit default swap (CDS) spreads is a challenging task. In this paper, we propose the Merton-LSTM model, a modified LSTM model formed by integrating with the Merton determinants model, to forecast the CDS indices. We provide the rigorous math behind the Merton-LSTM model, which demonstrates that by leveraging the nonlinear learning ability of LSTM with increased model capacity, the Merton-LSTM model is expected to learn the inherent association between the Merton determinants and CDS spreads. Further, the Merton-LSTM model is compared with the machine learning models LSTM, gated recurrent unit (GRU), multilayer perceptron network (MLP), support vector machine (SVM) and a typical stochastic series model in forecasting the two most liquid five-year CDS indices, North America High Yield index (CDX.NA.HY) and North America Investment Grade index (CDX.NA.IG) through the root mean squared error (RMSE) and the Diebold-Mariano test. The comparison results show that the RMSEs of the Merton-LSTM model are the lowest (6.2570–27.2000 for CDX.NA.HY and 1.3168–6.4772 for CDX.NA.IG) compared to other competitive models. The superiority of the Merton-LSTM model in forecasting performance is highlighted in long-term prediction even with a forecasting horizon extended to 28 days. Simulated trading with different thresholds and horizons is conducted in this study. We find that the Merton-LSTM trading strategy yields the highest annualized Sharpe ratios and lowest maximum losses at most thresholds and horizons, highlighting the economic significance of the proposed model. [ABSTRACT FROM AUTHOR]

This paper investigates the prediction of geological conditions ahead of tunnel boring machines (TBM) using a hybrid deep learning approach. By integrating graph convolutional network (GCN) and long short-term memory (LSTM) networks, the spatial and temporal features from TBM parameters and geological information are extracted for accurate prediction. The results from the case study indicate that (1) The proposed approach provides estimation with a high accuracy of 0.9986; (2) The past geological information has a significant contribution to the model; (3) The proposed approach outperforms several state-of-the-art methods including support vector machine (SVM), extreme gradient boosting (XGBoost) and LSTM method. The proposed hybrid deep learning approach can be a useful tool that provides reliable estimation of the advanced geological conditions in real-time. • A novel deep learning approach is proposed for advance geological perception. • The dependencies among the TBM parameters and geological information are considered. • A deep-learning framework that incorporates GCN and LSTM is established. • A realistic tunnel project in Singapore is used to demonstrate as a case study. • A high prediction accuracy can be achieved with average accuracy of 0.9986. [ABSTRACT FROM AUTHOR]

• A new deep-stacked autoencoder model is proposed for short-term load forecasting. • A stack of autoencoders is developed to extract more latent relevant features. • The challenge of setting the random initial weights of the deep models is solved. • The proposed model is equipped with multi-stage temporal attention mechanisms. • The proposed model has good robustness against various data partitioning. This paper presents an innovative univariate Deep LSTM-based Stacked Autoencoder (DLSTM-SAE) model for short-term load forecasting, equipped with a Multi-Stage Attention Mechanism (MSAM), including an input AM and several temporal AM in the pre-training phase. The input AM is used to capture the high-impact load sequence time steps of univariate input data. It should be noted that the model's performance is improved by increasing the network depth; however, finding the optimal network parameters is a challenging task due to the random assignment of the initial weights of the network. An unsupervised greedy layer-wise pre-training structure equipped with the MSAM is expanded to solve setting the random initial weight problem of the DLSTM-SAE model. The multi-stage temporal AM in the pre-training structure leads the DLSTM-SAE to properly learn the time dependencies related to remarkably long sequence input data and capture the temporal merit features lied in the LSTM memory. The performance of the proposed model is evaluated through various comparative tests with current prevalent models using actual energy market data New England ISO using three criteria indexes. The results show the superiority of the proposed model and its robustness in offline and online load forecasting. [ABSTRACT FROM AUTHOR]

The phase-resolved wave prediction of short crest waves is important to marine structures for both predicting deterministic motion and assisting decision-making. The present short crest wave phase-resolved prediction methods utilize Fast Fourier Transform (FFT) to process wave elevation fields in a large area, and these methods have shortcomings in calculation efficiency, calculation accuracy, and convenience. This paper proposes a long short-term memory wave prediction model (LSTM-WP model) based on deep learning to achieve a phase-resolved wave prediction of short crest waves. A tank experiment is conducted to verify and analyze the LSTM-WP model. From the results, it can be found that the LSTM-WP model provides high-precision predictions for the short crest wave surface under sea states of levels 4–7. Furthermore, the effects of the direction spectrum, predicted distance, and lead steps are discussed. It can be found that as the direction of the short crest wave becomes more concentrated, the prediction error rises more rapidly with increasing sea states. As the predicted distance increases, the prediction error of the LSTM-WP model increases linearly. As the number of lead steps increases, the prediction error of the LSTM-WP model shows a trend of first decreasing and then increasing. • A LSTM based wave model is developed for phase-resolved wave prediction of short crest wave. • Effect of environment for prediction performance are investigated. • Effect of lead step for prediction performance are investigated. • Verification studies are carried out using experimental datasets. • The LSTM based wave prediction model shows satisfactory performance in PRWP. [ABSTRACT FROM AUTHOR]

Predictive process analytics, often underpinned by deep learning techniques, is a newly emerged discipline dedicated for providing business process intelligence in modern organisations. Whilst accuracy has been a dominant criterion in building predictive capabilities, the use of deep learning techniques comes at the cost of the resulting models being used as 'black boxes', i.e., they are unable to provide insights into why a certain business process prediction was made. So far, little attention has been paid to interpretability in the design of deep learning-based process predictive models. In this paper, we address the 'black-box' problem in the context of predictive process analytics by developing attention-based models that are capable to inform both what and why is a process prediction. We propose i) two types of attentions— event attention to capture the impact of specific events on a prediction, and attribute attention to reveal which attribute(s) of an event influenced the prediction; and ii) two attention mechanisms— shared attention mechanism and specialised attention mechanism to reflect different design decisions between whether to construct attribute attention on individual input features (specialised) or using the concatenated feature tensor of all input feature vectors (shared). These lead to two distinct attention-based models, and both are interpretable models that incorporate interpretability directly into the structure of a process predictive model. We conduct experimental evaluation of the proposed models using real-life dataset and comparative analysis between the models for accuracy and interpretability, and draw insights from the evaluation and analysis results. The results demonstrate that i) the proposed attention-based models can achieve reasonably high accuracy; ii) both are capable of providing relevant interpretations (when validated against domain knowledge); and iii) whilst the two models perform equally in terms of prediction accuracy, the specialised attention-based model tends to provide more relevant interpretations than the shared attention-based model, reflecting the fact that the specialised attention-based model is designed to facilitate better interpretability. [ABSTRACT FROM AUTHOR]

Commercial buildings are a significant consumer of energy worldwide. Logistics facilities, and specifically warehouses, are a common building type which remain under-researched in the demand-side energy forecasting literature. Warehouses have an idiosyncratic profile when compared to other commercial and industrial buildings with a significant reliance on a small number of energy systems. As such, warehouse owners and operators are increasingly entering energy performance contracts with energy service companies (ESCOs) to minimise environmental impact, reduce costs, and improve competitiveness. ESCOs and warehouse owners and operators require accurate forecasts of their energy consumption so that precautionary and mitigation measures can be taken. This paper explores the performance of three machine learning models (Support Vector Regression (SVR), Random Forest, and Extreme Gradient Boosting (XGBoost)), three deep learning models (Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU)), and a classical time series model, Autoregressive Integrated Moving Average (ARIMA) for predicting daily energy consumption. The dataset comprises 8040 records generated over an 11-month period from January to November 2020 from a non-refrigerated logistics facility located in Ireland. The grid search method was used to identify the best configurations for each model. The proposed XGBoost models outperformed other models for both very short-term load forecasting (VSTLF) and short-term load forecasting (STLF); the ARIMA model performed the worst. [ABSTRACT FROM AUTHOR]

• Propose a novel hybrid model for multi-step ahead daily coal price forecasting. • Decompose original data into multiple modes via variational mode decomposition. • Use long short-term memory network containing attention layer to predict each mode. • Predicted mode results are ensembled by support vector regression model. • Proposed hybrid model is superior to baseline models in coal price forecasting. Accurate and reliable coal price prediction is of great significance to enhance the stability of the coal market. Numerous methods have been developed to improve the prediction performance. However, most of the studies adopt single model for coal price forecasting, and their accuracy and applicability are usually restricted. In this paper, we propose a novel hybrid VMD-A-LSTM-SVR model to achieve accurate multi-step ahead prediction of coal price. The proposed model consists of three valuable strategies. First, variational mode decomposition (VMD) decomposes the original coal price into several relatively regular sub modes to reduce the non-stationarity and uncertainty of the data. Second, the long short-term memory (LSTM) integrated with attention mechanism trains and predicts the decomposed modes individually to better capture the temporal information of historical data. Lastly, a support vector regression (SVR) model ensembles the predicted results of each mode into the final forecasted coal price. The experimental results of three typical coal price datasets demonstrate that the proposed strategies are all valuable for improving the forecasting performance. Moreover, the proposed model outperforms all state-of-the-art baseline models in terms of both model accuracy and stability. Extensive cross-comparisons of performance between models clearly indicate that the proposed hybrid algorithm is more effective and practical for coal price forecasting. [ABSTRACT FROM AUTHOR]

• Deep neural network models that jointly predict the next location and arrival time of a vehicle. • Learns sequential patterns in trajectory data using LSTM networks and self-attention mechanism. • Compares the effect of different model structures - hybrid LSTM and sequential LSTM. • Shows the importance of considering travel time in the next location prediction task. This paper aims to incorporate travel time prediction in the next location prediction problem to enable the prediction of the city-wide movement trajectory of an individual vehicle by considering both where the vehicle will go next and when it will arrive. We propose two deep learning models based on long short-term memory (LSTM) neural networks with self-attention mechanism—namely, hybrid LSTM and sequential LSTM. These models capture patterns in location and time sequences in trajectory data and their dependencies to predict next locations and travel times simultaneously. Using Bluetooth vehicle trajectory data from Brisbane, Australia, we compare the prediction performance of the proposed models with several existing approaches including hidden Markov model and other LSTM-based models. The results show that the proposed models produce higher prediction accuracy for both location and time prediction tasks, with the sequential LSTM yielding the best performance. Compared to the conventional next location prediction problem, which considers location sequences only without travel time consideration, the study finds that jointly modelling location and travel time sequences actually improves the next location prediction performance itself, potentially because travel time observations capture the information on traffic conditions in the network, which may affect drivers' location choices. We demonstrate an application of the proposed models in network traffic management, where important locations can be identified to mitigate congestion in a hot-spot by predicting where vehicles come from and go next in an urban road network. [ABSTRACT FROM AUTHOR]

• Introducing a new feature selection method based on entropy and mutual information to rank candidate inputs and eliminate redundant inputs according to their information value. • Using of wavelet transform to eliminate fluctuation behaviors of the electricity load and price time series decomposed the original signals. • Using of deep learning time series prediction based on LSTM to improve the forecast accuracy. Nowadays, a basic commodity for a human being to lead a standard lifestyle with human comfort irrespective of the nature of environmental conditions is electric power. The electricity load demand increases tremendously especially for a metropolitan city due to climatic conditions, population growth, local area development, industries expansion, air pollution, thermal device usage, etc. Hence, the accuracy of electricity load and its price forecasting is a deciding factor for the power distribution network to retain as an efficient, sustainable, and secure consumer-friendly network. On the other hand, based on the volatile, intermittent, and uncertain behavior of electricity load and price, an accurate, and robust forecast model should be designed. In this paper, a new hybrid forecast model for short-term electricity load and price prediction has been developed. The proposed method includes three modules: wavelet transform that is used to eliminate fluctuation behaviors of the electricity load and price time series, feature selection based on entropy and mutual information has been proposed to rank candidate inputs and eliminate redundant inputs according to their information value, and a new learning algorithm. The proposed learning method consists of a deep learning algorithm with LSTM networks which improves the accuracy of predictions. The performance of the proposed method has been validated successfully on load and price data collected from the Pennsylvania-New Jersey-Maryland (PJM) and Spain electricity markets. Also, for further test, the load data in Iran have been used. [ABSTRACT FROM AUTHOR]

Certain difficulties in path forecasting and filtering problems are based in the initial hypothesis of estimation and filtering techniques. Common hypotheses include that the system can be modeled as linear, Markovian, Gaussian, or all at one time. Although, in many cases, there are strategies to tackle problems with approaches that show very good results, the associated engineering process can become highly complex, requiring a great deal of time or even becoming unapproachable. To have tools to tackle complex problems without starting from a previous hypothesis but to continue to solve classic challenges and sharpen the implementation of estimation and filtering systems is of high scientific interest. This paper addresses the forecast–filter problem from deep learning paradigms with a neural network architecture inspired by natural language processing techniques and data structure. Unlike Kalman, this proposal performs the process of prediction and filtering in the same phase, while Kalman requires two phases. We propose three different study cases of incremental conceptual difficulty. The experimentation is divided into five parts: the standardization effect in raw data, proposal validation, filtering, loss of measurements (forecasting), and, finally, robustness. The results are compared with a Kalman filter, showing that the proposal is comparable in terms of the error within the linear case, with improved performance when facing non-linear systems. [ABSTRACT FROM AUTHOR]

Due to the intermittent nature of solar energy, accurate photovoltaic power predictions are very important for energy integration into existing energy systems. The evolution of deep learning has also opened the possibility to apply neural network models to predict time series, achieving excellent results. In this paper, a five layer CNN-LSTM model is proposed for photovoltaic power predictions using real data from a location in Temixco, Morelos in Mexico. In the proposed hybrid model, the convolutional layer acts like a filter, extracting local features of the data; then the temporal features are extracted by the long short-term memory network. Finally, the performance of the hybrid model with five layers is compared with a single model (a single LSTM), a CNN-LSTM hybrid model with two layers and two well known popular benchmarks. The results also shows that the hybrid neural network model has better prediction effect than the two layer hybrid model, the single prediction model, the Lasso regression or the Ridge regression. [ABSTRACT FROM AUTHOR]

• Deep Learning can be applied in financial forecast, improving results of classic models. • Proposed model with architecture based on a CNN-LSTM combination. • Images associated to time series that have static and dynamic information of the data. • CNN-LSTM model to predict the realized volatility for the price of gold. • The proposed model improves the results respect a normal LSTM network and GARCH. Prediction of volatility for different types of financial assets is one of the tasks of greater mathematical complexity in time series prediction, mainly due to its noisy, non-stationary and heteroscedastic structure. On the other hand, gold is an asset of particular importance for hedging and diversification of investment portfolios, and therefore it is important to predict future volatility of this asset. This paper seeks to significantly improve the forecast of gold volatility by combining two deep learning methodologies: short-term memory networks (LSTM) added to convolutional neural networks (specifically a pre-trained VGG16 network). It is important to mention that these types of hybrid architectures have not been used in time series prediction, so it is a completely new approach to solving these types of problems. The CNN-LSTM hybrid model is capable of including images as input which provides a wide variety of information associated with both static and dynamic characteristics of the series. In parallel, different lags of profitability of the series are entered as input, which allows it to learn from the temporal structure. The results show a substantial improvement when this hybrid model is compared to the GARCH and LSTM models. A 37% reduction in MSE is observed compared to the classic GARCH model, and 18% compared to the LSTM model. Finally, the Model Confidence Model (MCS) determines a significant improvement in the prediction of the hybrid model. The fundamental importance of this research lies in the application of a new type of architecture capable of processing various sources of information for any time series prediction task. [ABSTRACT FROM AUTHOR]

Various indoor positioning methods have been developed to solve the "last mile on Earth". Ultra-wideband positioning technology stands out among all indoor positioning methods due to its unique communication mechanism and has a broad application prospect. Under non-line-of-sight (NLOS) conditions, the accuracy of this positioning method is greatly affected. Unlike traditional inspection and rejection of NLOS signals, all base stations are involved in positioning to improve positioning accuracy. In this paper, a Long Short-Term Memory (LSTM) network is used while maximizing the use of positioning equipment. The LSTM network is applied to process the raw Channel Impulse Response (CIR) to calculate the ranging error, and combined with the improved positioning algorithm to improve the positioning accuracy. It has been verified that the accuracy of the predicted ranging error is up to centimeter level. Using this prediction for the positioning algorithm, the average positioning accuracy improved by about 62%. [ABSTRACT FROM AUTHOR]

Ship trajectory prediction is a key requisite for maritime navigation early warning and safety, but accuracy and computation efficiency are major issues still to be resolved. The research presented in this paper introduces a deep learning framework and a Gate Recurrent Unit (GRU) model to predict vessel trajectories. First, series of trajectories are extracted from Automatic Identification System (AIS) ship data (i.e., longitude, latitude, speed, and course). Secondly, main trajectories are derived by applying the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm. Next, a trajectory information correction algorithm is applied based on a symmetric segmented-path distance to eliminate the influence of a large number of redundant data and to optimize incoming trajectories. A recurrent neural network is applied to predict real-time ship trajectories and is successively trained. Ground truth data from AIS raw data in the port of Zhangzhou, China were used to train and verify the validity of the proposed model. Further comparison was made with the Long Short-Term Memory (LSTM) network. The experiments showed that the ship's trajectory prediction method can improve computational time efficiency even though the prediction accuracy is similar to that of LSTM. [ABSTRACT FROM AUTHOR]

With the development of online advertising technology, the accurate targeted advertising based on user preferences is obviously more suitable both for the market and users. The amount of conversion can be properly increased by predicting the user's purchasing intention based on the advertising Conversion Rate (CVR). According to the high-dimensional and sparse characteristics of the historical behavior sequences, this paper proposes a LSLM_LSTM model, which is for the advertising CVR prediction based on large-scale sparse data. This model aims at minimizing the loss, utilizing the Adaptive Moment Estimation (Adam) optimization algorithm to mine the nonlinear patterns hidden in the data automatically. Through the experimental comparison with a variety of typical CVR prediction models, it is found that the proposed LSLM_LSTM model can utilize the time series characteristics of user behavior sequences more effectively, as well as mine the potential relationship hidden in the features, which brings higher accuracy and trains faster compared to those with consideration of only low or high order features. [ABSTRACT FROM AUTHOR]

The ability to identify and accurately predict abnormal behavior is important for health monitoring systems in smart environments. Specifically, for elderly persons wishing to maintain their independence and comfort in their living spaces, abnormal behaviors observed during activities of daily living are a good indicator that the person is more likely to have health and behavioral problems that need intervention and assistance. In this paper, we investigate a variety of deep learning models such as Long Short Term Memory (LSTM), Convolutional Neural Network (CNN), CNN-LSTM and Autoencoder-CNN-LSTM for identifying and accurately predicting the abnormal behaviors of elderly people. The temporal information and spatial sequences collected over time are used to generate models, which can be fitted to the training data and the fitted model can be used to make a prediction. We present an experimental evaluation of these models performance in identifying and predicting elderly persons abnormal behaviors in smart homes, via extensive testing on two public data sets, taking into account different models architectures and tuning the hyperparameters for each model. The performance evaluation is focused on accuracy measure. [ABSTRACT FROM AUTHOR]