Search

In financial field, a robust IT system is of vital importance to ensure the smooth operation of financial transactions. However, many financial corporations still depend on operators to identify and eliminate the system failures when financial IT systems break down. This traditional operation method is time consuming and extremely inefficient. To improve the efficiency and accuracy of system failure detection and thereby reduce the impact of system failures on financial services, we propose a novel machine learning-based framework to predict the occurrence of system exceptions and failures in a financial IT system. In particular, we first extract rich information from system logs and eliminate noises in the data. Then the cleaned data is leveraged as the input of our proposed anomaly prediction framework which consists of three modules: key performance indicator data prediction module, anomaly identification module and severity classification module. Notably, we design a hierarchical architecture of alarm classifiers and try to alleviate the influence of class-imbalance problem on the overall performance. Empirically, the experimental results demonstrate the superior performance of our proposed method on a real-world financial IT system log data set. [ABSTRACT FROM AUTHOR]

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

Coronavirus disease 2019 (Covid-19) is a contagious pandemic illness characterized by severe acute respiratory syndrome. The daily rise of Covid-19 instances and fatalities has resulted in worldwide lockdowns, quarantines and social distancing. Researchers have been working incredibly to develop precisely focused strategies to warfare the Covid-19 pandemic. This study aims to develop a cyclical learning rate optimized stacked generalization computational models (CLR-SGCM) for predicting Covid-19 pandemic outbreaks. Stacked generalization framework performs hierarchical two-phase prediction. In the first phase, deep learning models namely Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) and statistical model Auto Regressive Integrated Moving Average (ARIMA) are used as sub models to create pooled datasets (PDS). Cyclical learning rate (CLR) optimizer is used to enhance learning rate of ensemble deep learning models namely LSTM and GRU. In the second phase, meta learner is trained on dataset PDS using four different regression algorithms such as linear regression, polynomial regression, lasso regression and ridge regression to perform the final predictions. Time series data from India, Brazil, and the United States were utilized to forecast the Covid-19 pandemic outbreak. According to experimental finding, the presented stacking ensemble model outpaces the individual learners in terms of accuracy and error rate. [ABSTRACT FROM AUTHOR]

The restored interest now receives renewable energy due to the global decline in greenhouse gas emanations and fossil fuel combustion. The fasted growing energy source, wind energy generation, is recognized as a clean energy source that has grown fast and is used extensively in wind power-producing facilities. This study's short-term wind speed estimations are made using a multivariate model based on an artificial neural network (ANN) that combines several local measurements, including wind speed, wind direction, LV active power, and theoretical power curve. The dataset was received from Turkey's SCADA system at 10-min intervals, and the actual data validated the expected performance. The research took wind speed into account as an input parameter and created a multivariate model. To perform prediction outcomes on time series data, an algorithm such as an artificial neural network (ANN) is utilized. The experiment verdicts reveal that the ANN algorithm produces reliable predicting results when metrics like 0.693 for MSE, 0.833 for RMSE and 0.96 for R-squared or Co-efficient of determination are considered. [ABSTRACT FROM AUTHOR]

The prediction of total solar irradiance (TSI) time series holds significant importance in the study of solar activity and the assessment of solar energy resources. The variational mode decomposition-based long short-term memory (VMD-LSTM) model is a hybrid deep learning model that demonstrates high prediction accuracy on long-term time series. To address the information leakage issue faced by hybrid models based on VMD and other data preprocessing methods, this study proposes a prediction method for hybrid deep learning models called dualfusion variational mode decomposition (DFVMD), which modifies the VMD decomposition approach. The DFVMD-LSTM model utilizes multiple TSI datasets as model features, and multisource datasets and multiple model comparison experiments are employed to verify the applicability and robustness of the model. The experimental results show that the DFVDM-LSTM model significantly reduces the periodic TSI prediction deviation introduced by the LSTM model. Furthermore, regardless of the training period or prediction horizon, the DFVMD-LSTM model exhibits an average root mean square error (RMSE) reduction of 14.79 % and an average mean absolute error (MAE) decrease of 21.50 %, demonstrating the superior predictive performance and improved reliability of the DFVDM-LSTM method. [ABSTRACT FROM AUTHOR]

Two stage grid search accepted as a promising heuristic search technique, involves a search performed in two stages. In the first stage a search is performed in coarse grain/low resolution to identify the optimal region and, in the second stage, a fine grain/high resolution search is performed in the neighborhood of the optimal region to identify the optimal parameters. Performing a search in two stages considerably reduces the computational complexity when compared to the basic grid search algorithm. However, an exhaustive search is to be carried out in the subspace during the second stage which may again be a computationally expensive task. The main contribution of this paper is to develop a new heuristic search technique which explores the discrete parameter space dimension wise recursively. The time complexity of the proposed algorithm is less than that of the two-stage grid search. The performance of the proposed algorithm in terms of required number of probes and time for optimal model selection, compared with the two-stage grid search, is verified for correctness and efficiency. [ABSTRACT FROM AUTHOR]

Sulphur dioxide is one of the most common air pollutants, forming acid rain and other harmful substances in the atmosphere, which can further damage our ecosystem and cause respiratory diseases in humans. Therefore, it is essential to monitor the concentration of sulphur dioxide produced in industrial processes in real-time to predict the concentration of sulphur dioxide emissions in the next few hours or days and to control them in advance. To address this problem, we propose an AR-LSTM analytical forecasting model based on ARIMA and LSTM. Based on the sensor's time series data set, we preprocess the data set and then carry out the modeling and analysis work. We analyze and predict the proposed analysis and prediction model in two data sets and conduct comparative experiments with other comparison models based on the three evaluation indicators of R2, RMSE and MAE. The results demonstrated the effectiveness of the AR-LSTM analytical prediction model; Finally, a forecasting exercise was carried out for emissions in the coming weeks using our proposed AR-LSTM analytical forecasting model. [ABSTRACT FROM AUTHOR]

Leaky Integrator Echo State Network (Leaky-ESN) is a useful training method for handling time series prediction problems. However, the singular coupling of all neurons in the reservoir makes Leaky-ESN less effective for sophisticated learning tasks. In this paper, we propose a new improvement to the Leaky-ESN model called the Multiple-Reservoir Hierarchical Echo State Network (MH-ESN). By introducing a new mechanism for constructing the reservoir, the efficiency of the network in handling training tasks is improved. The hierarchical structure is used in the process of constructing the reservoir mechanism of MH-ESN. The MH-ESN consists of multiple layers, each comprising a multi-reservoir echo state network model. The sub-reservoirs within each layer are linked via principal neurons, which mimics the functioning of a biological neural network. As a result, the coupling among neurons in the reservoir is decreased, and the internal dynamics of the reservoir are improved. Based on the analysis results, the MH-ESN exhibits significantly better prediction accuracy than Leaky-ESN for complex time series prediction. [ABSTRACT FROM AUTHOR]

Data-driven methods for computer simulations are blooming in many scientific areas. The traditional approach to simulating physical behaviors relies on solving partial differential equations (PDEs). Since calculating these iterative equations is highly both computationally demanding and time-consuming, data-driven methods leverage artificial intelligence (AI) techniques to alleviate that workload. Data-driven methods have to be trained in advance to provide their subsequent fast predictions; however, the cost of the training stage is non-negligible. This article presents a predictive model for inferencing future states of a specific fluid simulation that serves as a use case for evaluating different training alternatives. Particularly, this study compares the performance of only CPU, multi-GPU, and distributed approaches for training a time series forecasting deep learning model. With some slight code adaptations, results show and compare, in different implementations, the benefits of distributed GPU-enabled training for predicting high-accuracy states in a fraction of the time needed by the computational fluid dynamics solver. [ABSTRACT FROM AUTHOR]

The Echo State Network (ESN) is one of the most used machine learning methods for predicting chaotic time series. The topology of an ESN comprises three layers, namely: input, hidden, and output layer. In this work, different ESN topologies are analyzed by modifying the structure and number of internal connections in the hidden layer, which neurons are connected randomly. The ESN topologies are analyzed to have 2%, 6%, 20%, 40 % , ... , and up to 100% of internal connections. The prediction error is associated to the Mean Squared Error (MSE), which results have similar values for all ESN topologies, as long as the Echo state property is guaranteed. It is highlighted that the prediction horizon obtained with 2% of internal connections, is more significant than that obtained with 100%, providing a prediction of 80 and 60 steps-ahead, respectively. The prediction results significantly impact the implementation in a field-programmable gate array (FPGA), so that an enhanced hardware realization is introduced herein by reducing the use of multipliers by more than 90% compared with a fully connected topology, without significantly affecting the measure of the prediction horizon and MSE. Another contribution is the FPGA implementation of the activation function that is an hyperbolic tangent function for which different alternatives are shown to minimize the error of the approximated function without significantly increasing the implementation cost or latency. • FPGA implementation of the hyperbolic tangent function with reduced error. • Analysis of ESN topologies to verify chaotic time series prediction of large horizons. • Hardware reduction of the ESN reflected in about 90% of less multipliers. [ABSTRACT FROM AUTHOR]

In this paper, we present a feasibility study on predicting the torque signal of a passenger car with the help of a neural network. In addition, we analyze the possibility of using the proposed model structure for temperature prediction. This was carried out with a neural network, specifically a three-layer long short-term memory (LSTM) network. The data used were real road load data from a Jaguar Land Rover Evoque with a Twinster gearbox from GKN. The torque prediction generated good results with an accuracy of 55% and a root mean squared error (RMSE) of 49 Nm, considering that the data were not generated under laboratory conditions. However, the performance of predicting the temperature signal was not satisfying with a coefficient of determination (R2) score of −1.396 and an RMSE score of 69.4 °C. The prediction of the torque signal with the three-layer LSTM network was successful but the transferability of the network to another signal (temperature) was not proven. The knowledge gained from this investigation can be of importance for the development of virtual sensor technology.

Prediction with high reliability is very important in solving real-world problems, especially those that affect public health. The statistical properties of complex problems such as Covid-19 disease constantly change over time which makes modeling of such problems associated with high-level uncertainty. It has been proven that the type-2 fuzzy logic has the potential for modeling uncertainty to solve complex problems. In this research, for the first time, an intelligent method based on the capability of type-2 fuzzy logic was presented to manage uncertainty in predicting short-term and long-term time series in environmental crises such as the Covid-19 pandemic. The performance of the proposed model was evaluated using a real dataset collected from official sources. The results confirm the high efficiency of the proposed method on Covid-19 based on a ROC curve analysis. The obtained results showed an efficiency of 93.81% for short and 91.33% for long-term time series. This indicates the high efficiency and capability of the proposed model for managing uncertainty in predicting patterns of Covid-19 in comparison with similar methods. The proposed model can be useful to take strategic decisions and prevent the consequences of the Covid-19 epidemic in the short and long terms.

With the rapid development of meteorology, there requires a great need to better forecast dew point temperatures contributing to mild building surface and rational chemical control, while researches on time series forecasting barely catch the attention of meteorology. This paper would employ the seasonal-trend decomposition-based simplified dendritic neuron model (STLDNM*) to predict the dew point temperature. We utilize the seasonal-trend decomposition based on LOESS (STL) to extract three subseries from the original sequence, among which the residual part is considered as an input of an improved dendritic neuron model (DNM*). Then the back-propagation algorithm (BP) is used for training DNM* and the output is added to another two series disposed. Four datasets, which record dew points of four cities, along with eight algorithms are put into the experiments for comparison. Consequently, the combination of STL and simplified DNM achieves the best speed and accuracy.

Abstract In this work, a high efficient next generation reservoir computing (HENG‐RC) paradigm that adopts the principle of local states correlation and attention mechanism is proposed, which is able to process dynamical information generated by both the low dimensional and very large spatiotemporal chaotic systems (VLSCS). From a dynamical system perspective, the dynamical characteristics such as density distribution, Poincaré plots and max Lyapunov exponents of the proposed HENG‐RC are studied. It is revealed that the trained model can be seen as a data‐driven chaotic system. Furthermore, a novel scheme of secure communication based on chaotic synchronization of two HENG‐RC systems is designed, of which the security is enhanced as the intruder needs to know simultaneously the training signal and details of the parameter setting in the HENG‐RC. The digital implementation using field programmable gate array is experimentally realised, proving the feasibility of the secure communication scheme.

Traffic flow prediction is considered to be one of the fundamental technologies in intelligent transportation systems (ITSs) with a tremendous application prospect. Unlike traditional time series analysis tasks, the key challenge in traffic flow prediction lies in effectively modelling the highly complex and dynamic spatiotemporal dependencies within the traffic data. In recent years, researchers have proposed various methods to enhance the accuracy of traffic flow prediction, but certain issues still persist. For instance, some methods rely on specific static assumptions, failing to adequately simulate the dynamic changes in the data, thus limiting their modelling capacity. On the other hand, some approaches inadequately capture the spatiotemporal dependencies, resulting in the omission of crucial information and leading to unsatisfactory prediction outcomes. To address these challenges, this paper proposes a model called the Dynamic Spatial–Temporal Self-Attention Network (DSTSAN). Firstly, this research enhances the interaction between different dimension features in the traffic data through a feature augmentation module, thereby improving the model’s representational capacity. Subsequently, the current investigation introduces two masking matrices: one captures local spatial dependencies and the other captures global spatial dependencies, based on the spatial self-attention module. Finally, the methodology employs a temporal self-attention module to capture and integrate the dynamic temporal dependencies of traffic data. We designed experiments using historical data from the previous hour to predict traffic flow conditions in the hour ahead, and the experiments were extensively compared to the DSTSAN model, with 11 baseline methods using four real-world datasets. The results demonstrate the effectiveness and superiority of the proposed approach.

Detecting anomalies in large, complex systems is a critical and challenging task, and this is especially true for high-dimensional anomaly detection due to the underlying dependency structures among sensors. To incorporate the interrelationships among various sensors, a novel sparsity-constrained vector autoregressive moving average (scVARMA) model is proposed for anomaly detection in complex systems with multisensory signals. This model aims to leverage the inherent relationships and dynamics among various sensor readings, providing a more comprehensive and accurate analysis suitable for complex systems’ complex behavior. This research uses convex optimization to search for a parameterization that is sparse based on the principal of parsimony. This sparse model will not only contribute to meeting the real-time requirements of online monitoring strategies but also keeps the correlations among different sensory signals. The performance of the proposed scVARMA model is validated using real-world data from complex systems. The results affirm the superiority of the proposed scheme, demonstrating its enhanced performance and potential in practical applications.

Time series prediction is an effective tool for marine scientific research. The Hierarchical Temporal Memory (HTM) model has advantages over traditional recurrent neural network (RNN)-based models due to its online learning and prediction capabilities. Given that the neuronal structure of HTM is ill-equipped for the complexity of long-term marine time series applications, this study proposes a new, improved HTM model, incorporating Gated Recurrent Units (GRUs) neurons into the temporal memory algorithm to overcome this limitation. The capacities and advantages of the proposed model were tested and evaluated on time series data collected from the Xiaoqushan Seafloor Observatory in the East China Sea. The improved HTM model both outperforms the original one in short-term and long-term predictions and presents results with lower errors and better model stability than the GRU model, which is proficient in long-term predictions. The findings allow for the conclusion that the mechanism of online learning has certain advantages in predicting ocean observation data.

Electroencephalography (EEG) serves as a diagnostic technique for measuring brain waves and brain activity. Despite its precision in capturing brain electrical activity, certain factors like environmental influences during the test can affect the objectivity and accuracy of EEG interpretations. Challenges associated with interpretation, even with advanced techniques to minimize artifact influences, can significantly impact the accurate interpretation of EEG findings. To address this issue, artificial intelligence (AI) has been utilized in this study to analyze anomalies in EEG signals for epilepsy detection. Recurrent neural networks (RNNs) are AI techniques specifically designed to handle sequential data, making them well-suited for precise time-series tasks. While AI methods, including RNNs and artificial neural networks (ANNs), hold great promise, their effectiveness heavily relies on the initial values assigned to hyperparameters, which are crucial for their performance for concrete assignment. To tune RNN performance, the selection of hyperparameters is approached as a typical optimization problem, and metaheuristic algorithms are employed to further enhance the process. The modified hybrid sine cosine algorithm has been developed and used to further improve hyperparameter optimization. To facilitate testing, publicly available real-world EEG data is utilized. A dataset is constructed using captured data from healthy and archived data from patients confirmed to be affected by epilepsy, as well as data captured during an active seizure. Two experiments have been conducted using generated dataset. In the first experiment, models were tasked with the detection of anomalous EEG activity. The second experiment required models to segment normal, anomalous activity as well as detect occurrences of seizures from EEG data. Considering the modest sample size (one second of data, 158 data points) used for classification models demonstrated decent outcomes. Obtained outcomes are compared with those generated by other cutting-edge metaheuristics and rigid statistical validation, as well as results’ interpretation is performed.

With the large-scale development of electric vehicles, the accuracy of electric vehicle charging load prediction is increasingly important for electric power system. Accurate EV charging load prediction is essential for the efficiency of electric system planning and economic operation of electric system. This paper proposes an electric vehicle charging load predicting method based on variational mode decomposition and Prophet-LSTM. Firstly, the variational mode decomposition algorithm is used to decompose the charging load into several intrinsic mode functions in order to explore the characteristics of EV charging load data. Secondly, in order to make full use of the advantages of various forecasting methods, the intrinsic mode functions are classified into low and high frequency sequences based on their over-zero rates. The high and low frequency sequences are reconstructed to obtain two frequency sequences. Then the LSTM neural network and Prophet model are used to predict the high and low frequency sequences, respectively. Finally, the prediction results obtained from the prediction of high frequency and low frequency sequences are combined to obtain the final prediction result. The assessment of the prediction results shows that the prediction accuracy of the prediction method proposed in this paper is improved compared to the traditional prediction methods, and the average absolute error is lower than that of ARIMA, LSTM and Prophet respectively by 7.57%, 8.73%, and 46.02%. The results show that the prediction method proposed in this paper has higher prediction accuracy than the traditional methods, and is effective in predicting EV charging load.

Spiking neural networks (SNNs) are considered the most promising new generation of artificial neural networks, due to their superior dynamic structures and low energy consumption, resembling that of the biological brain. Recent studies have suggested that SNNs could benefit from online learning in dynamic scenarios involving temporal sequences. However, the network performance of traditional spiking encoding methods is significantly affected by noise. As such, this study proposes a quantum-inspired online spiking neural network (QiSNN), which combines a quantum particle swarm optimization algorithm and a Kalman filtering technique to smooth and denoise the original time-series data. Additionally, a novel adaptive threshold selection method is developed to determine the similarity between neurons in a repository. The resulting model is applied to a dataset from the Department of Environment, Food, and Rural Affairs (DEFRA) in the UK, and used to predict ozone and PM10 concentrations that characterize air quality. Experimental results demonstrate that the proposed QiSNN outperforms baseline models across multiple evaluation metrics. [ABSTRACT FROM AUTHOR]

In recent years, how to discover causality rather than correlation among climate variables and how to use causality to help in time-series tasks have received great concern. However, the high dimensionality and nonlinearity of climate variables are the main issues for causal inference based on historical large-scale climate time series. Therefore, a method based on neural Granger causality inference is proposed to study the interactions of climate variables, with focus on the variables commonly used in the energy field, especially in photovoltaics. Firstly, for each climate variable, the time-varying causality and global causality are, respectively, obtained on each time window and on the whole series by neural Granger causality inference. Secondly, the global causality is used as a feature selection map of the input variables in the prediction task. Finally, compared with some existing feature selection methods, the experiments determine that the proposed method not only reveals the appropriate causality rather than the correlation among climate variables, but also efficiently reduces the input dimensionality and improves the performance and interpretability of the predicting model. [ABSTRACT FROM AUTHOR]

The piston ring pack is one of the most important components to affect the power, economy and emission of a diesel engine. Among them, lube oil consumption (LOC) and crankcase blow-by (CB) are the main indicators to evaluate the lubrication and sealing performance of the piston ring pack. In this study, the multi-factor analysis was implemented to predict the influencing factors on the LOC, CB and the sensitivity between them in the piston ring pack for an agricultural diesel engine. A YNF40 diesel engine was taken as the research object. Firstly, the test bench of LOC and CB was built for the diesel engine. The kinematics and dynamics models were established using hydrodynamic lubrication and gas flow theoretical of piston pack friction subsets, technical and structural parameters. A series of experimental tests were also carried out to verify the accuracy and reliability of the simulation. Secondly, the grey relational grade (GRG) of different influencing factors was obtained by grey relation analysis (GRA). The sensitivity analysis between the influencing factors and LOC-CB was then carried out using GRG calculation and ranking. Finally, the time-series prediction models were established for the transient variations of LOC and CB under different speed conditions using a back propagation (BP) neural network. The multi-factor GRG analysis indicated that the most sensitive influencing factor on the LOC and CB in the piston ring pack for a diesel engine was the chamfer of the bottom edge of the 1st ring groove of the piston, whereas, the most insensitive influencing factor was the back clearance of the 3rd ring groove of piston. The maximum GRG between LOC and CB and the chamfer of the bottom edge of the 1st ring groove of the piston was 0.892 79, whereas, the minimum GRG between LOC and CB and the back clearance of the 3rd ring groove of the piston was 0.583 61. The descending order of sensitivity was obtained between the target and influencing factors. The LOC and CB gradually increased in the piston ring pack for a diesel engine, as the chamfer of the bottom edge of the 1st ring groove of the piston increased. The back clearance of the 3rd ring groove of the piston shared no significant effect. The average standard deviation of the variation of LOC and CB with the chamfer of the bottom edge of the 1st ring groove of the piston were 2.81, and 3.90, respectively. By contrast, the average standard deviation of the variation of LOC and CB with the back clearance of the 3rd ring groove of the piston were 0.003 and 0.209, respectively. Therefore, the larger the GRG between LOC and CB, and each influencing factor were, the larger the mean standard deviation of LOC and CB with the influencing factors were, and the more sensitive the objectives were to the influencing factors. The time-series prediction showed that there was a better consistent trend of the predicted and actual values on the transient variations of LOC and CB. The prediction accuracy and reliability of the time series prediction model were higher than before, which were distributed within the 95% confidence band. The research findings can provide effective guidance to improve the thermal efficiency with the less carbon emissions of agricultural diesel engines. [ABSTRACT FROM AUTHOR]

Accurately predicting severe accident data in nuclear power plants is of utmost importance for ensuring their safety and reliability. However, existing methods often lack interpretability, thereby limiting their utility in decision making. In this paper, we present an interpretable framework, called GRUS, for forecasting severe accident data in nuclear power plants. Our approach combines the GRU model with SHAP analysis, enabling accurate predictions and offering valuable insights into the underlying mechanisms. To begin, we preprocess the data and extract temporal features. Subsequently, we employ the GRU model to generate preliminary predictions. To enhance the interpretability of our framework, we leverage SHAP analysis to assess the contributions of different features and develop a deeper understanding of their impact on the predictions. Finally, we retrain the GRU model using the selected dataset. Through extensive experimentation utilizing breach data from MSLB accidents and LOCAs, we demonstrate the superior performance of our GRUS framework compared to the mainstream GRU, LSTM, and ARIMAX models. Our framework effectively forecasts trends in core parameters during severe accidents, thereby bolstering decision-making capabilities and enabling more effective emergency response strategies in nuclear power plants. [ABSTRACT FROM AUTHOR]

Long Short-Term Memory (LSTM) Neural Network has great potential to predict sequential data. Time series prediction is one of the most popular experimental subjects of LSTM. To this end, various LSTM algorithms have been developed to predict time series data. However, there are a few works considering the hyperparameter optimization of LSTM along with parallelization approaches. To address this problem, a parallelized classic LSTM is proposed to predict time series. In the preprocessing phase, it first replaces missing values with zero and then normalizes the time series matrix. The transposed matrix is divided into training and testing parts. Consequently, a core-based parallelism is established, thereby utilizing forking to split prediction into multiple processes. Derivative-free optimization techniques are also analyzed to find out what sort of hyperparameter optimization technique is much more feasible for a parallelized LSTM. Further, a state-of-the-art comparison is included in the study. Experimental results show that training loss is optimal when using Nelder–Mead. Employing effort-intensive optimization methods such as genetic algorithms results in a remarkable CPU time reduction in parallelized designs according to the results. Last, the proposed algorithm outperformed the comparison methods with respect to the prediction errors. [ABSTRACT FROM AUTHOR]

Ali Network data based on the Qinghai-Tibetan Plateau (QTP) can provide representative coverage of the climate and surface hydrometeorological conditions in the cold and arid region of the QTP. Among them, the plateau soil moisture can effectively quantify the uncertainty of coarse resolution satellite and soil moisture models. With the objective of constructing an "end-to-end" soil moisture prediction model for the Tibetan Plateau, a combined prediction model based on time series decomposition and a deep neural network is proposed in this article. The model first performs data preprocessing and seasonal-trend decomposition using loess (STL) to obtain the trend component, seasonal component and random residual component of the original time series in an additive way. Subsequently, the bidirectional gated recurrent unit (BiGRU) is used for the trend component, and the long short-term memory (LSTM) is used for the seasonal and residual components to extract the time series information. The experiments based on the measured data demonstrate that the use of STL decomposition and the combination model can effectively extract the information in soil moisture series using its concise and clear structure. The proposed model in this article has a stable performance improvement of 5-30% over a single model and existing prediction models in different prediction time domains. In long-range prediction, the proposed model also achieves the best accuracy in the shape and temporal domains described by using dynamic time warping (DTW) index and temporal distortion index (TDI). In addition, the generalization performance experiments show that the combined method proposed in this article has strong reference value for time series prediction of natural complex systems. [ABSTRACT FROM AUTHOR]

General aviation accidents have complex interactions and influences within them that cannot be simply explained and predicted by linear models. This study is based on chaos theory and uses general aviation accident data to conduct research on different timescales (HM-scale, ET-scale, and EF-scale). First, time series are constructed by excluding seasonal patterns from the statistics of general aviation accidents. Secondly, the chaotic properties of multi-timescale series are determined by the 0–1 test and Lyapunov exponent. Finally, by introducing the sparrow search algorithm and tent chaotic mapping, a CSSA-LSSVM prediction model is proposed. The accident data of the National Transportation Safety Board (NTSB) of the United States in the past 15 years is selected for case analysis. The results show that the phase diagram of the 0–1 test presents Brownian motion characteristics, and the maximum Lyapunov exponents of the three scales are all positive, proving the chaotic characteristics of multi-timescale series. The CSSA-LSSVM prediction model's testing results illustrate its superiority in time series predicting, and when the timescale declines, the prediction error reduces gradually while the fitting effect strengthens and then decreases. This study uncovers the nonlinear chaotic features of general aviation accidents and demonstrates the significance of multi-timescale research in time series analysis and prediction. [ABSTRACT FROM AUTHOR]

The geometric precision of machined parts is greatly reduced due to thermal errors, thereby the thermal errors should be controlled and compensated timely to improve the machining accuracy. But the previously established thermal error models in related studies are not robust, and the predictive accuracy is not high, and the convergence is weak. Moreover, the error control delay is serious due to the network bandwidth limitation. To overcome the above challenges, the theoretical and data-driven methods are combined. The intrinsic long-term memory characteristic of thermal errors is revealed by the theoretical method and the finite element analysis, providing a theoretical guidance for the data-driven method. Then a residual-based one-dimensional convolution-minimum gate unit model is designed based on the residual connection. The designed one-dimensional convolution layer extracts the data features, and the residual connection with multiple pooling layers is used to compress the dimension of the error data. The minimal gate unit is used to improve the convergence rate. Then a new physical-data-edge-cloud terminal architecture is proposed by combining the advantages of the cloud computing and edge computing. Finally, the experiment was conducted to verify the designed physical-data-edge-cloud terminal architecture. The results indicate that the predictive accuracy is 98.18% and that the convergence time is shortened to 13.719 s compared with traditional thermal error models. Furthermore, the geometric error is reduced by more than 80%, and the total time consumption is 166 s, and the data transmission time is 25 s. The system efficiency of the designed physical-data-edge-cloud terminal system is far higher than that of physical-edge-cloud, physical-data-cloud, and physical-cloud systems. [ABSTRACT FROM AUTHOR]

Concerns that impair human societies frequently include a heavy dependence on petroleum and coal and emissions of greenhouse gases. Thus, adopting renewable energy sources, such as wind power, has become a practical solution to this problem. Therefore, to carry out the research on wind velocity energy, a time-series structure is necessary. This study uses the Markov chain Monte Carlo approach and the Seasonal Autoregressive Integrated Moving Average (SARIMA) model to estimate short-term and long-term sustained winds. The significance of building a wind energy system is initially discussed, after which a wind velocity time-series framework based on a SARIMA is presented, followed by a short-term and Long-term wind speed projection. Furthermore, a methodology utilizing the Markov chain Monte Carlo method (MCMC) is suggested to establish a wind energy time-series analysis. This framework draws a Markov chain for time-series data on wind energy to maintain stochasticity and realize the probability transition matrix. Gibbs sampling is employed as well. The model's forecasting abilities were tested using the original database and various efficiency assessment measures, including Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) with efficiency of 13.09 and 1.03. In this study, a framework with the highest KGE and WI as well as the lowest RMSE and MAE was chosen. The findings demonstrate that the approach used in the operation provides outstanding predictability. [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]

This paper focuses on water quality prediction in the presence of a large number of missing values in water quality monitoring data. Current water quality monitoring data mostly come from different monitoring stations in different water bodies. As the duration of water quality monitoring increases, the complexity of water quality data also increases, and missing data is a common and difficult to avoid problem in water quality monitoring. In order to fully exploit the valuable features of the monitored data and improve the accuracy of water quality prediction models, we propose a long short-term memory (LSTM) encoder-decoder model that combines a Kalman filter (KF) with an attention mechanism. The Kalman filter in the model can quickly complete the reconstruction and pre-processing of hydrological data. The attention mechanism is added between the decoder and the encoder to solve the problem that traditional recursive neural network models lose long-range information and fully exploit the interaction information among high-dimensional covariate data. Using original data from the Haimen Bay water quality monitoring station in the Lianjiang River Basin for analysis, we trained and tested our model using detection data from 1 January 2019 to 30 June 2020 to predict future water quality. The results show that compared with traditional LSTM models, KF-LSTM models reduce the average absolute error ( M A E ) by 10%, the mean square error ( M S E ) by 21.2%, the root mean square error ( R M S E ) by 13.2%, while increasing the coefficient of determination ( R 2 ) by 4.5%. This model is more suitable for situations where there are many missing values in water quality data, while providing new solutions for real-time management of urban aquatic environments. [ABSTRACT FROM AUTHOR]

At present, time series prediction methods are widely applied for Human–Computer Intelligent Systems in various fields such as Finance, Meteorology, and Medicine. To enhance the accuracy and stability of the prediction model, this paper proposes a time series prediction method called RAdam-Dual stage Attention mechanism-Nested Long Short-Term Memory (RAdam-DA-NLSTM). First, we design a Nested LSTM (NLSTM), which adopts a new internal LSTM unit structure as the memory cell of LSTM to guide memory forgetting and memory selection. Then, we design a self-encoder network based on the Dual stage Attention mechanism (DA-NLSTM), which uses the NLSTM encoder based on the input attention mechanism, and uses the NLSTM decoder based on the time attention mechanism. Additionally, we adopt the RAdam optimizer to solve the objective function, which dynamically selects Adam and SGD optimizers according to the variance dispersion and constructs the rectifier term to fully express the adaptive momentum. Finally, we use multiple datasets, such as PM2.5 data set, stock data set, traffic data set, and biological signals, to analyze and test this method, and the experimental results show that RAdam-DA-NLSTM has higher prediction accuracy and stability compared with other traditional methods. [ABSTRACT FROM AUTHOR]

Sea surface temperature is an important parameter of ocean hydrology. Accurate prediction of sea surface temperature is of great significance for ocean economic development and extreme weather prevention. Application of deep learning-based method in sea surface temperature prediction has significantly increased due to its high analytical power. Nevertheless, sea surface temperature time series are so volatile and stochastic, leading to the fact that in-depth analysis and accurate prediction of sea surface temperature are still challenging. Considering non-stationary and nonlinear characteristics in sea surface temperature sequence, variational mode decomposition is adopted as the de-noising module to reduce the influence of noise. Furthermore, the convolutional neural network is combined with the gated recurrent unit network to extract both spatial and temporal features in sea surface temperature sequence, simultaneously. Finally, a sea surface temperature prediction model based on deep learning model with frequency analysis module is proposed in this paper. The sea surface temperature of East China Sea is selected for empirical study. Comparative analysis demonstrated that the proposed model significantly outperforms the existing sea surface temperature prediction models. Especially, even to the mesoscale level, mean absolute error and root mean squared error of the proposed method are still 0.2038 and 0.2481, respectively. [ABSTRACT FROM AUTHOR]

To effectively improve the power dispatching, the prediction accuracy of wind power has been the concern of many scholars for many years. The wind power prediction problem is actually equivalent to the wind speed prediction problem. Based on linear regression (LR) and variational mode decomposition (VMD), in this paper, we proposed an efficient hybrid method to predict wind speed. In the proposed method, the VMD is used to decompose the signal of wind speed into several sub-signal. Compared with the original wind-speed series, each sub-signal is a more stable subsequence signal. Then, we used the LR method to predict each subsequence signal. Eventually, we obtain the final prediction results of the original wind speed series merged the forecasting values of all subsequences signal. We selected two data to test our proposed method in our experiment. Compared with several comparison methods, we found that our proposed methods has better prediction performance than other methods from the experimental results. [ABSTRACT FROM AUTHOR]

n recent years, there has been significant progress in the fields of Attention-based Transformers[1, 2]. In this paper, we compare the performance of an Attention encoder and an LSTM[3] for the stock prediction task, evaluating the performance based on the model's size and input length. We utilize stock history data for Apple, Samsung, and Amazon, obtained from Kaggle. The Attention Encoder model outperforms the LSTM baseline model on all datasets. This demonstrates the superior performance of the Attention Encoder model. Additionally, increasing the input sequence dimension of the Attention Encoder network leads to improved performance. However, there is a concern regarding overfitting of the Attention Encoder model due to the rapid changes in the data. [ABSTRACT FROM AUTHOR]

Time series forecasting using multilayer feed-forward neural networks (FNN) is potential to give high accuracy. Several factors influence the accuracy. One of them is the choice of activation functions (AFs). There are several AFs commonly used in FNN with their specific characteristics, such as bounded type AFs. They include sigmoid, softsign, arctan, and tanh. This paper investigates the effect of the amplitude in the bounded AFs on the FNNs' accuracy. The theoretical investigations use simplified FNN models: linear equation and linear combination. The results show that the higher amplitudes give higher accuracy than typical amplitudes in softsign, arctan, and tanh AFs. However, in sigmoid AF, the amplitude changes do not influence the accuracy. These theoretical results are supported by experiments using the FNN model for time series prediction of 10 foreign exchanges from different continents compared to the US dollar. Based on the experiments, the optimum amplitude of the AFs should be high, that is greater or equal to 100 times of the maximum input values to the FNN, and the accuracy gains up to 3--10 times. [ABSTRACT FROM AUTHOR]

In this paper, we investigate the capability of modeling distant temporal interaction of Long Short-Term Memory (LSTM) and introduce a novel Long Short-Term Memory on time series problems. To increase the capability of modeling distant temporal interactions, we propose a hierarchical architecture (HLSTM) using several LSTM models and a linear layer. This novel framework is then applied to electric power consumption, real-life crime and financial data. We demonstrate in our simulations that this structure significantly improves the modeling of deep temporal connections compared to the classical architecture of LSTM and various studies in the literature. Furthermore, we analyze the sensitivity of the new architecture with respect to the hidden size of LSTM. [ABSTRACT FROM AUTHOR]

The prediction of wind in terms of both wind speed and direction, which has a crucial impact on many real-world applications like aviation and wind power generation, is extremely challenging due to the high stochasticity and complicated correlation in the weather data. Existing methods typically focus on a sub-set of influential factors and thus lack a systematic treatment of the problem. In addition, fine-grained forecasting is essential for efficient industry operations, but has been less attended in the literature. In this work, we propose a novel data-driven model, multi-horizon spatiotemporal network (MHSTN), generally for accurate and efficient fine-grained wind prediction. MHSTN integrates multiple deep neural networks targeting different factors in a sequence-to-sequence (Seq2Seq) backbone to effectively extract features from various data sources and produce multi-horizon predictions for all sites within a given region. MHSTN is composed of four major modules. First, a temporal module fuses coarse-grained forecasts derived by numerical weather prediction (NWP) and historical on-site observation data at stations so as to leverage both global and local atmospheric information. Second, a spatial module exploits spatial correlation by modeling the joint representation of all stations. Third, an ensemble module weighs the above two modules for final predictions. Furthermore, a covariate selection module automatically choose influential meteorological variables as initial input. MHSTN is already integrated into the scheduling platform of one of the busiest international airports of China. The evaluation results demonstrate that our model outperforms competitors by a significant margin. [ABSTRACT FROM AUTHOR]

Accurately predicting traffic flow on roads is crucial to address urban traffic congestion and save on travel time. However, this is a challenging task due to the strong spatial and temporal correlations of traffic data. Existing traffic flow prediction methods based on graph neural networks and recurrent neural networks often overlook the dynamic spatiotemporal dependencies between road nodes and excessively focus on the local spatiotemporal dependencies of traffic flow, thereby failing to effectively model global spatiotemporal dependencies. To overcome these challenges, this article proposes a new Spatio-temporal Causal Graph Attention Network (STCGAT). STCGAT utilizes a node embedding technique that enables the generation of spatial adjacency subgraphs on a per-time-step basis, without requiring any prior geographic information. This obviates the necessity for intricate modeling of constantly changing graph topologies. Additionally, STCGAT introduces a proficient causal temporal correlation module that encompasses node-adaptive learning, graph convolution, as well as local and global causal temporal convolution modules. This module effectively captures both local and global Spatio-temporal dependencies. The proposed STCGAT model is extensively evaluated on traffic datasets. The results show that it outperforms all baseline models consistently. [ABSTRACT FROM AUTHOR]

Data concerning product sales are a popular topic in time series forecasting due to their multidimensionality and wide presence in many businesses. This paper describes the research in predicting the timing and product category of the next purchase based on historical customer transaction data. Given that the dataset was acquired from a vendor of medical drugs and devices, the generic product identifier (GPI) classification system was incorporated in assigning product categories. The models built are based on recurrent neural networks (RNN) and long short-term memory (LSTM) neural networks with different input and output features, and training datasets. Experiments with various datasets were conducted and optimal network structures and types for predicting both product category and next purchase day were identified. The key contribution of this research is the process of data transformation from its original purchase transaction format into a time series of input features for next purchase prediction. With this approach, it is possible to implement a dedicated personalized marketing system for a vendor. [ABSTRACT FROM AUTHOR]

Collaborative robots have gained popularity in industries, providing flexibility and increased productivity for complex tasks. However, their ability to interact with humans and adapt to their behavior is still limited. Prediction of humanmovement intentions is one way to improve the robots adaptation. This paper investigates the performance of using Transformers and MLP-Mixer based neural networks to predict the intended human arm movement direction, based on gaze data obtained in a virtual reality environment, and compares the results to using an LSTM network. The comparison will evaluate the networks based on accuracy on several metrics, time ahead of movement completion, and execution time. It is shown in the paper that there exists several network configurations and architectures that achieve comparable accuracy scores. The best performing Transformers encoder presented in this paper achieved an accuracy of 82.74%, for predictions with high certainty, on continuous data and correctly classifies 80.06% of the movements at least once. The movements are, in 99% of the cases, correctly predicted the first time, before the hand reaches the target and more than 19% ahead of movement completion in 75% of the cases. The results shows that there are multiple ways to utilize neural networks to perform gaze based arm movement intention prediction and it is a promising step toward enabling efficient human-robot collaboration. [ABSTRACT FROM AUTHOR]

• A new grey intelligent prediction algorithm with the multiobjective correction strategy is proposed. • A new weakened accumulation operation is proposed. • A new evaluation system for evaluating the performance of prediction models is constructed. As a combination of the differential equation prediction model and intelligent optimization algorithm, the grey intelligent prediction algorithm has attracted increasing attention due to its outstanding performance in small-sample environments. However, most studies focus only on the improvement of algorithm performance, with little regard to the uniformity, ill-condition and overfitting of the algorithm. To promote the development of this field, we develop a new grey intelligent prediction algorithm with multiobjective correction strategy based on the new weakened accumulation grey optimization model and the multiobjective grey wolf optimizer. In this new prediction algorithm, the new weakened accumulation operation is utilized to enhance the predictive ability of the algorithm and mitigate the ill-condition of the system, the Bernoulli parameter and a discretization technique are used to activate the uniformity and unbiasedness of the algorithm, respectively, and the multiobjective grey wolf optimizer is employed to alleviate the overfitting of the system. Compared with the previous grey intelligent prediction models, the new prediction algorithm is more perfect and reasonable. Taking two energy datasets as research cases, the evaluation results of a system consisting of six evaluation metrics and the Diebold-Mariano test show that the proposed model outperforms the other six comparative models in terms of prediction performance and stability, which confirms the feasibility and validity of the algorithm. [ABSTRACT FROM AUTHOR]

One of the main indicators for detecting changes in climate and marine ecosystems around the world is the sea surface temperature (SST). Even with several models presented in the literature, it is still a challenging task when only a single model is considered for SST forecasting. In this context, hybrid approaches that combine statistical models (to estimate linear dependencies) and machine learning models (to estimate nonlinear dependencies from residuals) have attained highlighted accuracy in several time series forecasting problems. In this way, this paper proposes a hybrid system that combines the autoregressive integrated moving average (ARIMA) model with a deep morphological neural network model, which employs dilation-erosion operators to estimate ARIMA's model residuals. In this context, the proposed hybrid system is composed of three stages: (1) time series forecast using the ARIMA model, (2) residuals forecast using the deep morphological neural network and (3) a linear combination of the stages (1) and (2). Three SST time series are used in the experimental analysis. Moreover, the achieved results are evaluated using three relevant statistical measures showing that the proposed hybrid system attains higher accuracy when compared to other hybrid models in the literature. [ABSTRACT FROM AUTHOR]

Forecasting container throughput is critical for improved port planning, operations, and investment strategies. Reliability of forecasting methods need to be ensured before utilizing their outcomes in decision making. This study compares forecasting performances of various time series methods, namely autoregressive integrated moving average (ARIMA), seasonal ARIMA (SARIMA), Holt-Winter's Exponential Smoothing (HWES), and the Prophet model. Since forecast combinations can improve performance, simple and weighted combinations of ARIMA, SARIMA and HWES have been explored, too. Monthly container throughput data of port of Shanghai, Busan, and Nagoya are used. The Prophet model outperforms others in the in-sample forecasting, while combined models outperform others in the out-sample forecasting. Due to the observed differences between the in-sample and out-sample forecast accuracy measures, this study proposes a forecast performance metric consistency check approach for informed real-world applications of forecasting models in port management decision-making. [ABSTRACT FROM AUTHOR]

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

In this paper, we present a feasibility study on predicting the torque signal of a passenger car with the help of a neural network. In addition, we analyze the possibility of using the proposed model structure for temperature prediction. This was carried out with a neural network, specifically a three-layer long short-term memory (LSTM) network. The data used were real road load data from a Jaguar Land Rover Evoque with a Twinster gearbox from GKN. The torque prediction generated good results with an accuracy of 55% and a root mean squared error ( R M S E ) of 49 Nm, considering that the data were not generated under laboratory conditions. However, the performance of predicting the temperature signal was not satisfying with a coefficient of determination ( R 2 ) score of −1.396 and an R M S E score of 69. 4 °C. The prediction of the torque signal with the three-layer LSTM network was successful but the transferability of the network to another signal (temperature) was not proven. The knowledge gained from this investigation can be of importance for the development of virtual sensor technology. [ABSTRACT FROM AUTHOR]

Tensor time series shows the co-evolution characteristics of multiple time series with time changes, and tensor time series prediction has become an important problem. Aiming at the high-dimensional characteristics of tensor time series, T-Transformer, a tensor time series prediction method based on Transformer, is proposed, which integrates tensor operation and Transformer into a unified framework. The method first uses tensors to represent higher-order time series, uses tensor slices and vectorization to convert tensor time series into vectors, and then inputs the vectors into Transformer model after coding, and finally obtains the predicted value of tensor time series. Experiments on three open data sets show that the proposed method achieves better prediction results than the benchmark method. [ABSTRACT FROM AUTHOR]

Artificial Neural Network (ANN) is an effective machine learning technique for addressing regression tasks. Nonetheless, the performance of ANN is highly dependent on the values of its parameters, specifically the weight and bias. To improve its predictive generalization, it is crucial to optimize these parameters. In this study, the Barnacles Mating Optimizer (BMO) is employed as an optimization tool to automatically optimize these parameters. As a relatively new optimization algorithm, it has been shown to be effective in addressing various optimization problems. The proposed hybrid predictive model of BMO-ANN is tested on time series data of stock price using six selected inputs to predict the next day’ closing prices. Evaluated based on Mean Square Error (MSE) and Root Mean Square Error (RMSPE), the proposed BMO-ANN exhibits significant superiority over the other identified hybrid algorithms. Additionally, the difference in means between BMO-ANN and other identified hybrid algorithms was found to be statistically significant, with a significance level of 0.05%. [ABSTRACT FROM AUTHOR]

Nowadays, air pollution is a big environmental problem in developing countries. In this problem, particulate matter 2.5 (PM2.5) in the air is an air pollutant.When its concentration in the air is high in developing countries like Vietnam, it will harm everyone's health. Accurate prediction of PM2.5 concentrations can help to make the correct decision in protecting the health of the citizen. This study develops a hybrid deep learning approach named PM25-CBL model for PM2.5 concentration prediction in Ho Chi Minh City, Vietnam. Firstly, this study analyzes the effects of variables on PM2.5 concentrations in Air Quality HCMC dataset. Only variables that affect the results will be selected for PM2.5 concentration prediction. Secondly, an efficient PM25-CBL model that integrates a convolutional neural network (CNN) and Bidirectional Long Short-TermMemory (Bi-LSTM) is developed. This model consists of three following modules: CNN, Bi-LSTM, and Fully connected modules. Finally, this study conducts the experiment to compare the performance of our approach and several state-of-the-art deep learning models for time series prediction such as LSTM, Bi-LSTM, the combination of CNN and LSTM (CNN-LSTM), and ARIMA. The empirical results confirm that PM25-CBL model outperforms other methods for Air Quality HCMC dataset in terms of several metrics including Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). [ABSTRACT FROM AUTHOR]

Nowadays, China is developing new energy industries to reduce carbon emissions to meet the challenge of world climate change, so investors can consider to invest in stocks of Chinese new energy companies to gain income. In order to study how to forecast stock closing prices of new energy companies in China, we have chosen 12 representative companies, and first used autoregressive univariate time series models to predict the trends of the stock closing prices in the next month. The results show that Seasonal Autoregressive Integrated Moving Average model has the best out-of-sample trend prediction effect. Second, we use multivariate time series forecasting models to predict the stock closing prices of each day through external variables. The results show that Temporal Convolutional Attention Neural Networks has the best effect of out-of-sample prediction. We recommend that investors who are interested in investing in new energy companies in China first use the Seasonal Autoregressive Integrated Moving Average model to predict the short-term stock closing price trend in the future, and then use the Temporal Convolutional Attention Neural Networks model to predict the stock closing price on the next day to decide whether to invest.