Hierarchical Complementary Enhanced Autoencoder Integrating Spatio‐Temporal Interaction Feature for Soft Sensor

A novel deep learning prediction model for concrete dam displacements using interpretable mixed attention mechanism

A data-driven hybrid framework for voltage transformer ratio error prediction: Addressing challenges in complex power systems

Timely and accurate traffic flow information is essential for the realization of intelligent transportation systems (ITS). For the existing traffic flow prediction models, only the temporal features were extracted but the spatial features were neglects. A short-term traffic flow prediction model based on Pearson Correlation Coefficient (PCC) and Bi-directional Long Short-Term Memory Network (BiLSTM) model was proposed. The PCC was used to mine the spatial dependence of the road network, and the temporal features of traffic flow were mined using BiLSTM model. The extracted temporal and spatial features were combined to achieve short-term flow prediction. The experimental results show that the PCC-BiLSTM model has higher prediction accuracy and lower error than the prediction models considering only temporal features, which verifies the effectiveness of traffic flow prediction in consideration of both temporal and spatial characteristics.

Line loss accounts for most of the energy loss in low‐voltage distribution network. Accurate prediction of line loss rate (LLR) is of great significance to eliminate abnormal transmission line faults in time and ensure power supply safety. The existing prediction methods for LLR seldom consider the seasonal trend data that affect it, and there is a hysteresis effect in predicting non‐stationary line loss sequences. To solve the above problems, this paper proposes a prediction method based on multi‐dimensional information matrix and dimensional attention mechanism‐long‐and short‐term time‐series network (DAM‐LSTNet). Firstly, the maximum information coefficient (MIC) method is used to screen the distribution network characteristics and seasonal trend parameters. Secondly, the variational mode decomposition (VMD) method optimized by the genetic algorithm is used to decompose the historical line loss data and form a multi‐dimensional information matrix with the screened characteristic parameters of the network. Finally, the multi‐dimensional information matrix is put into the LSTNet network with dimensional attention mechanism to predict the LLR. The example analysis shows that compared with the existing methods, it has the advantages of sufficient consideration of prediction parameters, adaptive changes of prediction weights, weak hysteresis effect and high prediction accuracy.

Accurate prediction of wind turbine power is an important means to ensure the stable operation of wind turbines. The traditional wind speed-power curve only shows the relationship between wind turbine power and wind speed, without considering other influencing factors, such as rotor speed and pitch angle, and has certain limitations. Therefore, based on the supervisory control and data acquisition data of a wind farm, this paper proposes a hybrid multivariate prediction model that combines polynomial regression and random forest (RF) to predict wind turbine output power. First, to reduce the difficulty of expert analysis, the maximum information coefficient is used to analyze the correlation between data and select features. Wind speed, rotor speed, pitch angle, and wind direction are considered important and chosen for power prediction modeling. Then the relationship between multiple variables and output power is established through polynomial regression. Finally, the new polynomial features and output power are used to train the RF model to predict the output power. The experimental results show that the predicted wind speed and power diagram shows a band-like distribution, which matches closely with the real one; compared with multiple models such as a single RF model, a polynomial regression model, and a feedforward artificial neural network model, the polynomial-RF model has the highest prediction accuracy. The mean absolute percentage error (MAPE) of the polynomial-RF model prediction is 0.06, and the root mean square error (RMSE) is 114. Compared with the single RF model, the MAPE predicted by the polynomial-RF model is reduced by 14%, and the RMSE is reduced by 6%. It can be seen that the polynomial-RF model proposed in this paper has a good application prospect in the prediction of wind turbine output power.

Pedestrian trajectory prediction is fundamental to a wide range of scientific research work and industrial applications. Most of the current advanced trajectory prediction methods incorporate context information such as pedestrian neighbourhood, labelled static obstacles, and the background scene into the trajectory prediction process. In contrast to these methods which require rich contexts, the method in our paper focuses on predicting a pedestrian's future trajectory using his/her observed part of the trajectory only. Our method, which we refer to as LVTA, is a Location-Velocity-Temporal Attention LSTM model where two temporal attention mechanisms are applied to the hidden state vectors from the location and velocity LSTM layers. In addition, a location-velocity attention layer embedded inside a tweak module is used to improve the predicted location and velocity coordinates before they are passed to the next time step. Extensive experiments conducted on three large benchmark datasets and comparison with eleven existing trajectory prediction methods demonstrate that LVTA achieves competitive prediction performance. Specifically, LVTA attains 9.19 pixels Average Displacement Error (ADE) and 17.28 pixels Final Displacement Error (FDE) for the Central Station dataset, and 0.46 metres ADE and 0.92 metres FDE for the ETH&UCY datasets. Furthermore, evaluation on using LVTA to generate trajectories of different prediction lengths and on new scenes without the need of retraining confirms that it has good generalizability.

BackgroundA person's breathing pattern can be a reflection of their emotional and physical well-being because it shows the frequency, intensity, and rhythm of their breathing.ObjectiveThis research article presents a comprehensive approach to breathe pattern classification utilizing gyroscope and accelerometer readings obtained from individuals using two distinct sensors. The study encompasses the acquisition of six diverse breathing patterns, with a focus on data pre-processing through Min-Max normalization.MethodsTo select essential features from the normalized data, an innovative optimization algorithm, Adaptive Chimp Optimization (AdCO), is introduced. AdCO integrates an adaptive weighting strategy into the conventional Chimp optimization algorithm, enhancing convergence rates and enabling global optimal feature selection. Furthermore, the article introduces the application of the selected features in breath pattern classification using a hybrid deep learning mechanism, DABiG. DABiG leverages the Bidirectional Gated Recurrent Unit (BiGRU), a neural network architecture capable of processing sequential data bi-directionally.ResultsSpatial and temporal attention mechanisms are incorporated into DABiG to enhance its ability to focus on relevant spatial regions and time steps within the breath pattern data.ConclusionSpatial attention assigns weights to spatial regions, while temporal attention assigns weights to time steps, improving feature extraction and classification accuracy.

Spatial correlation learning based on graph neural network for medium-term wind power forecasting

GAOformer: An adaptive spatiotemporal feature fusion transformer utilizing GAT and optimizable graph matrixes for offshore wind speed prediction

Spiking neural networks (SNNs) possess great potential for energy-efficient computation; however, their practical deployment is often constrained by the high computational cost associated with training across multiple time steps. Unlike previous approaches that simply reduce the number of time steps, this work investigates temporal feature redundancy and reveals the feature overlapping phenomenon in SNNs-showing that substantial computational redundancy exists across temporal dimensions. To address this issue, we propose our core contribution, temporal differential decoupling (TDD), which transforms network computation into the differential domain to disentangle static and dynamic feature components. This decoupling enables focused processing of informative signals and significantly reduces redundant computation without compromising model accuracy. Building upon this framework, we further design the TDD-based differential domain low-sparsity approximation (TDD-DDLA) algorithm, which is based on the gradient sensitivity criterion, as an implementation strategy to quantify each temporal feature's contribution to gradient updates and achieve efficient energy optimization. In contrast to prior studies that only adjusted time steps without explicit feature-level analysis, our framework provides a structured and theoretically grounded analysis of temporal feature evolution. Experimental results demonstrate that our method achieves up to 80.9% fewer spikes per time step and 57.8% fewer total spikes, with no degradation in classification performance, offering a promising pathway toward scalable, low-cost, and high-accuracy SNNdeployment.

CircRNA is a non-coding RNA that has been considered as a marker of disease and can even be used as a drug target. Exploring the associations between circRNA and disease is beneficial for revealing the mechanisms of disease pathogenesis. Most of the existing computational-based methods do not consider the interaction in multiple features of circRNAs or diseases. Unlike traditional linear models, factorization machines take into account the interaction between features. In this study, we propose a method to Identify the CircRNA-Disease associations based on Feature Graph Factorization (ICDFGF) with graph attention mechanism. After extracting the sequence features, expression features, regulatory features of circRNAs and disease symptom features, disease-miRNA features, disease-gene features, an interactive network of features was constructed and learned by graph factorization. Compared with other existing prediction methods, ICDFGF has higher prediction accuracy. At the same time, the study also analyzes the predicted circRNA-disease associations, which can be corroborated by other studies.

With the vigorous development of social media, microblog and other social media have gained massive number of users. Hot events spread explosively, and the popularity prediction of microblog content has important research significance in politics, economy and people's livelihood. The existing models related to microblog popularity prediction adopts temporal sequence features or text features, which is relatively single, easily leads to the loss of useful information, and the prediction accuracy of the model is not high. Therefore, this paper designs a popularity prediction model TTMPP that combines temporal sequence features and text features. First, extracting the time series features based on the random point process model SEISMIC, and then the decision tree model lightGBM is used to predict the popularity of microblog based on temporal sequence features and text features. Finally, in the Sina Weibo data set, compared with a variety of typical popularity prediction models, experiments show that the model proposed in this paper has higher prediction accuracy.

Landslide displacement prediction has garnered significant recognition as a pivotal component in realizing successful early warnings and implementing effective control measures. This task remains challenging as landslide deformation involves not only temporal dependency within time series data but also spatial dependence across various regions within landslides. The present study proposes a landslide spatiotemporal displacement forecasting model by introducing attention-based deep learning algorithms based on spatiotemporal analysis. The Maximal Information Coefficient (MIC) approach is employed to quantify the spatial and temporal correlations within the daily data of Global Navigation Satellite System (GNSS) observations. Based on the quantitative spatiotemporal analysis, the proposed prediction model combines a convolutional neural network (CNN) and long short-term memory (LSTM) network to capture spatial and temporal dependencies individually. Spatial–temporal attention mechanisms are implemented to optimize the model. Additionally, we develop a single-point prediction model using LSTM and a multiple-point prediction model using the CNN-LSTM without an attention mechanism to compare the forecasting capabilities of the attention-based CNN-LSTM model. The Outang landslide in the Three Gorges Reservoir Area (TGRA), characterized by a large and active landslide equipped with an advanced monitoring system, is taken as a studied case. The temporal MIC results shed light on the response times of monitored daily displacement to external factors, showing a lagging duration of between 10 and 50 days. The spatial MIC results indicate mutual influence among different locations within the landslide, particularly in the case of nearby sites experiencing significant deformation. The attention-based CNN-LSTM model demonstrates an impressive predictive performance across six monitoring stations within the Outang landslide area. Notably, it achieves a remarkable maximum coefficient of determination (R2) value of 0.9989, accompanied by minimum values for root mean squared error (RMSE), absolute mean error (MAE), and mean absolute percentage error (MAPE), specifically, 1.18 mm, 0.99 mm, and 0.33%, respectively. The proposed model excels in predicting displacements at all six monitoring points, whereas other models demonstrate strong performance at specific individual stations but lack consistent performance across all stations. This study, involving quantitative deformation characteristics analysis and spatiotemporal displacement prediction, holds promising potential for a more profound understanding of landslide evolution and a significant contribution to reducing landslide risk.

MSST-ResNet: Deep multi-scale spatiotemporal features for robust visual object tracking

In this paper we present a hardware-assisted rendering technique coupled with a compression scheme for the interactive visual exploration of time-varying scalar volume data. A palette-based decoding technique and an adaptive bit allocation scheme are developed to fully utilize the texturing capability of a commodity 3-D graphics card. Using a single PC equipped with a modest amount of memory, a texture capable graphics card, and an inexpensive disk array, we are able to render hundreds of time steps of regularly gridded volume data (up to 45 millions voxels each time step) at interactive rates, permitting the visual exploration of large scientific data sets in both the temporal and spatial domain.