Temporal Prediction Method Research Articles

Study on Fast Temporal Prediction Method of Flame Propagation Velocity in Methane Gas Deflagration Experiment Based on Neural Network

To address the challenges of high experimental costs, complexity, and time consumption associated with pre-mixed combustible gas deflagration experiments under semi-open space obstacle conditions, a rapid temporal prediction method for flame propagation velocity based on Ranger-GRU neural networks is proposed. The deflagration experiment data are employed as the training dataset for the neural network, with the coefficient of determination (R2) and mean squared error (MSE) used as evaluation metrics to assess the predictive performance of the network. First, 108 sets of pre-mixed methane gas deflagration experiments were conducted, varying obstacle parameters to investigate methane deflagration mechanisms under different conditions. The experimental results demonstrate that obstacle-to-ignition source distance, obstacle shape, obstacle length, obstacle quantity, and thick and fine wire mesh obstacles all significantly influence flame propagation velocity. Subsequently, the GRU neural network was trained, and different activation functions (Sigmoid, Relu, PReLU) and optimizers (Lookahead, RAdam, Adam, Ranger) were incorporated into the backpropagation updating process of the network. The training results show that the Ranger-GRU neural network based on the PReLU activation function achieves the highest mean R2 value of 0.96 and the lowest mean MSE value of 7.16759. Therefore, the Ranger-GRU neural network with PReLU activation function can be a viable rapid prediction method for flame propagation velocity in pre-mixed methane gas deflagration experiments under semi-open space obstacle conditions.

TSAGNN: Temporal link predict method based on two stream adaptive graph neural network

Temporal link prediction based on graph neural networks has become a hot spot in the field of complex networks. To solve the problems of the existing temporal link prediction methods based on graph neural networks do not consider the future time-domain features and spatial-domain features are limited used, this paper proposes a novel temporal link prediction method based on two streams adaptive graph neural networks. Firstly, the network topology features are extracted from the micro, meso, and middle perspectives. Combined with the adaptive mechanism of convolution and self-attention, the preprocessing of the feature extraction is more effective; Secondly, an extended bi-directional long short-term memory network is proposed, which uses graph convolution to process topological features, and recursively learns the state vectors of the target snapshot by using the future time-domain information and the past historical information; Thirdly, the location coding is replaced by the time-coding for the transformer mechanism, so that past information and future information can be learned from each other, and the time-domain information of the network can be further mined; Finally, a novel two-stream network framework is proposed, which combines the processing results of point features and edge features. The experimental results on 9 data sets show that the proposed method has a better prediction effect and better robustness than the classical graph neural network methods.

Implementation of Long Short-Term Memory (LSTM) Networks for Stock Price Prediction

In this research, we explore the potential of Long Short-Term Memory (LSTM) networks for predicting stock prices. Due to the complexities of the financial markets and the inherent volatility of stock prices, accurate forecasting is now crucial for investors and financial specialists. It has been shown that LSTM, a type of recurrent neural network (RNN), can recognise temporal correlations and patterns in serial data. Training and assessing LSTM models in this work involves analysing stock price data, relevant financial measures, and market sentiment indicators. We looked into other ideas, hyper parameters, and preprocessing methods to see if we might boost the networks' performance. To further improve the model's generalizability, we utilise series normalisation and removal to reduce overfitting. The outcomes demonstrate that the LSTM network outperforms more standard series temporal prediction methods in capturing and anticipating shifts in action pricing. We also conduct extensive back testing and evaluation, using measures like mean squared error (MSE) and mean absolute error (MAE), to assess the model's accuracy and resilience. The results of this study shed light on how deep learning techniques, in particular LSTM networks, can be applied to the prediction of stock prices, potentially assisting traders, investors, and other financial decision-makers in navigating complex and volatile financial markets.

Predicting Temporal Sets with Simplified Fully Connected Networks

Given a sequence of sets, where each set contains an arbitrary number of elements, temporal sets prediction aims to predict which elements will appear in the subsequent set. Existing methods for temporal sets prediction are developed on sophisticated components (e.g., recurrent neural networks, attention or gating mechanisms, and graph neural networks), which inevitably increase the model complexity due to more trainable parameters and higher computational costs. Moreover, the involved nonlinear activation may contribute little or even degrade the performance. In this paper, we present a succinct architecture that is solely built on the Simplified Fully Connected Networks (SFCNs) for temporal sets prediction to bring both effectiveness and efficiency together. In particular, given a user's sequence of sets, we employ SFCNs to derive representations of the user by learning inter-set temporal dependencies, intra-set element relationships, and intra-embedding channel correlations. Two families of general functions are introduced to preserve the permutation-invariant property of each set and the permutation-equivariant property of elements in each set. Moreover, we design a user representations adaptive fusing module to aggregate user representations according to each element for improving the prediction performance. Experiments on four benchmarks show the superiority of our approach over the state-of-the-art under both transductive and inductive settings. We also theoretically and empirically demonstrate that our model has lower space and time complexity than baselines. Codes and datasets are available at https://github.com/yule-BUAA/SFCNTSP.

Dynamic change, driving mechanism and spatiotemporal prediction of the normalized vegetation index: a case study from Yunnan Province, China

Vegetation indexes have been widely used to qualitatively and quantitatively evaluate vegetation cover and its growth vigor. To further extend the study of vegetation indexes, this paper proposes to study the spatial and temporal distribution characteristics and specific driving mechanisms of vegetation indexes based on the example of Yunnan Province, China, and also adds the study of spatial and temporal prediction methods of vegetation indexes. This paper used data on this region’s normalized vegetation index (NDVI), three meteorological factors, and eight social factors from 1998 to 2019. The dynamic change in and driving mechanism of the NDVI were studied using mean value analysis, univariate linear trend regression analysis, and partial correlation analysis. In addition, the Fourier function model and the CA–Markov model were also used to predict the NDVI of Yunnan Province from 2020 to 2030 in time and space. The results show that: (1) The NDVI value in Yunnan Province is high, showing a significant growth trend. The increased vegetation coverage area has increased in the past 22 years without substantial vegetation degradation. (2) The positive promotion of meteorological factors is greater than the negative inhibition. The partial correlation of relative humidity among meteorological factors is the highest, which is the main driving factor. (3) The NDVI value is significantly positively correlated with population and economy and negatively correlated with pasture land and agricultural area. (4) The NDVI values are predicted well in time (R = 0.64) and space (Kappa = 0.8086 and 0.806), satisfying the accuracy requirements. This paper aims to enrich the theoretical and technical system of ecological environment research by studying the dynamic change, driving mechanism, and spatiotemporal prediction of the normalized vegetation index. Its results can provide the necessary theoretical basis for the simulation and prediction of vegetation indexes.

Temporal prediction for spectrum environment maps with moving radiation sources

AbstractSpectrum resources are becoming harder to come by for wireless communications. The spectrum environment map (SEM), which depicts the electromagnetic environment's current state and future trend, is a valuable technique for managing and allocating spectrum resources. Most SEM construction approaches only take static SEMs into account and cannot forecast time‐domain changes and trends of SEMs in dynamic scenes. In this paper, a brand‐new temporal SEM prediction method for the high dynamic spectrum environment is proposed. This method is based on knowledge of radiation source and the optical flow driven by propagation channel models. First, a novel radiation source localization strategy is designed to obtain the radiation source movement information. Then, the optical flow field of the available SEMs is combined with the information regarding radiation source movement. In order to forecast future SEMs, a propagation model driven reconstruction technique is developed. Simulation findings demonstrate how well the suggested strategy is tailored to capture the spatiotemporal correlation of SEMs. This technique performs better than the state‐of‐the‐art in terms of single‐ and multiple‐step SEM predictions.

A self-supervised temporal temperature prediction method based on dilated contrastive learning

Due to the scarcity of the labeled data, traditional supervised learning methods have a limited application scope, which caused the supervised-based model performance will greatly be decreased. In this paper, we propose a promising model based on self-supervised learning. To update the weight and the contrastive relation in the features, a new self-supervised loss, is introduced. First, the convolution neural network is used in the proposed network to extract the deep feature in the first processing. Second, the self-supervised long–short time memory (LSTM) sequential is constructed for further deal. At last, the teacher net and student net have coordinately fine-tuned the credibility of the temperature prediction. By the experimental comparison, our proposed CNN-SSDLSTM is competitive with other supervised and semi-supervised methods. The evaluation experiments achieve state-of-the-art performance in aluminum electrolysis temperature prediction applications.

A temporal and spatial prediction method for urban pipeline network based on deep learning

Water pipeline is one of the important components of urban infrastructure and plays a key role in residential life. An accurate pressure prediction could help improve the resilience of the system. In recent years, some studies have found that the massive pressure monitoring data have complex temporal and spatial correlations. It issues some new challenges to traditional prediction models. In this study, a pressure prediction method based on spatial–temporal neural network (PP-STNN) is proposed. Before the modeling, the pipeline network is mapped into a graph. In the method, Graph Convolutional Network (GCN) is used to capture the spatial correlation of the pipeline network and Gated Recurrent Unit (GRU) is used to capture the temporal correlation. The proposed method is evaluated using real-world dataset and compared with some benchmark methods. The results show that the proposed method could reach the highest accuracy among all methods for different prediction steps. Moreover, the comparison indicates that simultaneously considering temporal and spatial correlation can contribute to the prediction, especially for multiple steps prediction. Compared with GRU for 3-step, 12-step, and 24-step prediction, the proposed method can improve the Root Mean Square Error (RMSE) by about 19%, 8%, and 8%, respectively. Using deep learning methods, this study can improve the accuracy of the pressure prediction, thus increasing the resilience of the cities and promoting safety and sustainable development in the area.

SE-GRU: Structure Embedded Gated Recurrent Unit Neural Networks for Temporal Link Prediction

Temporal link prediction on dynamic graphs is essential to various areas such as recommendation systems, social networks, and citation analysis, and thus attracts great attention in both research and industry fields. For complex graphs in real-world applications, although recent temporal link prediction methods perform well in predicting high-frequency and nearby connections, it becomes more challenging when considering low-frequency and earlier connections. In this work, we introduce a novel and elegant prediction architecture called Structure Embedded Gated Recurrent Unit (SE-GRU) neural networks, to strengthen the prediction robustness against frequency variation and occurrence delay of connections. The established SE-GRU embeds the structure for local topological characteristics to emphasize the different connection frequencies between nodes and captures the temporal dependencies to avoid losing valuable information caused by long-term changes. We realize neural network optimization considering three terms concerning reconstruction, structure, and evolution. The extensive experiments performed on three public datasets demonstrate the significant superiority of SE-GRU compared with 5 representative and state-of-the-art competitors under three evaluation metrics. The results validate the effectiveness and robustness of our proposed method, by showing that the frequencies and timestamps of connections have a little-to-no negative impact on prediction accuracy.

Graph regularized nonnegative matrix factorization for link prediction in directed temporal networks using PageRank centrality

Temporal link prediction aims to predict new links at time T + 1 based on a given temporal network from time 1 to T. Some existing temporal link prediction methods only consider the naturally links direction formation but ignore the local and global information of temporal networks. To address this issue, in this paper, we propose a novel graph regularized nonnegative matrix factorization algorithm for the problem of temporal link prediction by using PageRank centrality and asymmetric link clustering coefficient, referred to as the GNMFCA. Specifically, we use graph regularization to capture the local information of each slice in temporal networks and PageRank centrality to compute the importance of nodes in each slice, which captures the global information of each slice in temporal networks. By jointly optimizing them in a nonnegative matrix factorization model, the proposed model can preserve both the local and global information simultaneously. Besides, we propose effective multiplicative updating rules to solve the proposed model and provide the convergence analysis of the iterative algorithm. Finally, we perform numerical experiments on eight artificial and four real-world temporal networks to demonstrate that GNMFCA outperforms some existing temporal link prediction algorithms.

TLP-CCC: Temporal Link Prediction Based on Collective Community and Centrality Feature Fusion.

In the domain of network science, the future link between nodes is a significant problem in social network analysis. Recently, temporal network link prediction has attracted many researchers due to its valuable real-world applications. However, the methods based on network structure similarity are generally limited to static networks, and the methods based on deep neural networks often have high computational costs. This paper fully mines the network structure information and time-domain attenuation information, and proposes a novel temporal link prediction method. Firstly, the network collective influence (CI) method is used to calculate the weights of nodes and edges. Then, the graph is divided into several community subgraphs by removing the weak link. Moreover, the biased random walk method is proposed, and the embedded representation vector is obtained by the modified Skip-gram model. Finally, this paper proposes a novel temporal link prediction method named TLP-CCC, which integrates collective influence, the community walk features, and the centrality features. Experimental results on nine real dynamic network data sets show that the proposed method performs better for area under curve (AUC) evaluation compared with the classical link prediction methods.

Testing collection‐time reduction in fine‐root biomass estimation in Atlantic Forests

AbstractAimsFine roots are essential components of the below‐ground layer and play an important role in the carbon cycle. Methods for root extraction and biomass estimation have been proposed, including the temporal prediction method. However, there are doubts if the best model to estimate total root mass varies between study sites. Additionally, there are no records regarding the prediction method's efficiency for shorter collection times than 40 min. Here, we aim to clarify these doubts.LocationBrazilian Atlantic Forest.MethodsWe extracted 1080 fine‐root samples from two contrasting ecosystems at 60 time intervals of 2 min each. We then performed a model selection to identify the best‐fit model and used it to find the shortest time suitable for collecting fine‐root samples (40, 32, 24, 16, or 8 min). A further 448 root samples were collected from seven ecosystems by employing the shortest time tested (8 min). We calculated the percentage of estimated mass at 120 min and tested for differences between ecosystems.ResultsWe found that Weibull was the best‐fit model, and it performed well for modeling root extraction at shorter collection times. All collection times tested had excellent goodness of fit, and there was strong evidence that the estimated mass did not differ between them. Moreover, collections at 8 min were enough to make reliable estimates of fine‐root mass at 120 min in all ecosystems.ConclusionsWeibull is a flexible model and can accurately estimate fine‐root mass at 120 min in different ecosystems. The extraction of fine roots can be reduced to four time intervals of 2 min each when using the temporal prediction method. By reducing the time spent removing roots from each soil core, researchers can increase the number of soil cores extracted per study site and characterize the environment properly.

Spatial—Temporal Traffic Flow Data Restoration and Prediction Method Based on the Tensor Decomposition

As an important part of urban big data, traffic flow data play a critical role in traffic management and emergency response. Traffic flow data contain multi-mode characteristics, which need to be deeply mined. To make full use of multi-mode characteristics, we use a 3-order tensor to represent the traffic flow data, considering “temporal-spatial-periodic” characteristics. To recover the missing data of traffic flow, we propose the Missing Data Completion Algorithm Based on Residual Value Tensor Decomposition (MDCA-RVTD), which combines linear regression, univariate spline, and CP decomposition. Then, we predict the future traffic flow data by using the proposed Traffic Flow Prediction Algorithm Based on Data Completion Strategy (TFPA-DCS). The experimental results show that recovering the missing data is helpful in improving the prediction accuracy. Additionally, the prediction accuracy of the proposed Algorithm is better than gray model and traditional tensor CP decomposition method.

A Novel Spatiotemporal Method for Predicting Covid-19 Cases

Prediction methods are important for many applications. In particular, an accurate prediction for the total number of cases for pandemics such as the Covid-19 pandemic could help medical preparedness by providing in time a sucient supply of testing kits, hospital beds and medical personnel. This paper experimentally compares the accuracy of ten prediction methods for the cumulative number of Covid- 19 pandemic cases. These ten methods include three types of neural networks and extrapola- tion methods based on best fit quadratic, best fit cubic and Lagrange interpolation, as well as an extrapolation method proposed by the second author. We also consider the Kriging and inverse distance weighting spatial interpolation methods. We also develop a novel spatiotemporal prediction method by combining temporal and spatial prediction methods. The experiments show that among these ten prediction methods, the spatiotemporal method has the smallest root mean square error and mean absolute error on Covid-19 cumulative data for counties in New York State between May and July, 2020.

A comparative study of different kinematics - based methods to predict the time of slope failure

Landslide forecasting is considered one of the key components of early warning systems. Predicting landslide failure at the slope-scale is a major scientific challenge, but on the other hand, it can mitigate the consequences of slope failures in terms of both human lives and economic losses. Therefore, landslide forecasting is a subject worthy of further research efforts due to its social implications. Landslide forecasting consists of the prediction of a slope failure in spatial and/or temporal terms. Temporal prediction can be performed at a regional/global scale or on a slope-scale. This paper focuses on the different methods for temporal prediction of landslides on a slope scale based on kinematic parameters such as displacement and its derivatives (velocity and acceleration). These kinematic parameters are directly related to the stability conditions of the moving mass. This paper accurately explains the correlation between the kinematics and the collapse time of a slope. Also, the present study systematically explains and compares the methods such as the empirical and semi-empirical methods for forecasting the failure time of slopes with their corresponding advantages and limitations. Finally, a detailed outline of the future technology and risks in this area are also introduced in this paper.

Spatio-temporal AI inference engine for estimating hard disk reliability

This paper focuses on building a spatio-temporal AI inference engine for estimating hard disk reliability. Most electronic systems such as hard disks routinely collect such reliability parameters in the field to monitor the health of the system. Changes in parameters as a function of time are monitored and any observed changes are compared with the known failure signatures. If the trajectory of the measured data matches that of a failure signature, operators are alerted to take corrective action. However, the interest of the operators lies in being able to identify the failures before they occur. The state of the art methodology including our prior work is to train machine learning models on temporal sequence data capturing the variations across multiple features and using it to predict the remaining useful life of the devices. However, as we show in this paper temporal prediction capability alone is not sufficient and can lead to low precision and the uncertainty around the prediction is very large. This is primarily due to the non-uniform progression of feature patterns over time. Our hypothesis is that the accuracy can be improved if we combine the temporal prediction methods with a spatial analysis that compares the value of key SMART features of the devices across similar model in a fixed time window (unlike the temporal method which uses the data from a single device and a much larger historical window). In this paper, we first describe both temporal and spatial approaches, describe the methods to select various hyperparameters, and then show a workflow to combine these two methodologies and provide comparative results. Our results illustrate that the average precision of temporal methods using long-short temporal memory networks to predict impending failures in the next ten days was 84 percent. To improve precision, we use the set of disks identified as potential failures and start applying spatial anomaly detection methods on those disks. This helps us remove the false positives from the temporal prediction results and provide a tighter bound on the set of disks with impending failure.

A new temporal prediction method of grazing pressure based on normalized difference vegetation index and precipitation using nonlinear autoregressive with exogenous input networks

AbstractRestoration of natural vegetation in arid and semi‐arid grasslands is facing severe challenges. The vegetation is easy to lose their vitality, resulting in the loss of the cover in natural grasslands under the high grazing pressure. To address this situation, this paper proposes a novel method for accurately predicting the grazing pressure using the nonlinear autoregressive with exogenous input (NARX) network based on the remote sensing data of normalized difference vegetation index (NDVI) and precipitation. The proposed method uses the NARX networks to predict the temporal variations of the NDVI with respect to the precipitation. The grazing pressure can be thus calculated using the predicted values of the NDVI. For practical application, this study investigated an arid and semi‐arid grassland with heavy grazing pressure in Hulunbuir, China. The results demonstrate that the proposed method can provide an accurate prediction of the grazing pressure (mean absolute error 0.103, root‐mean‐square error 0.122, mean absolute percentage error 8.36% and coefficient of determination 0.899 at the confidence interval of 95%). In addition, the predicted values of the grazing pressure in the study area during the years from 2016 to 2020 can be obtained using the proposed method. The proposed method can obtain a good prediction of the grazing pressure, which can be further used as a guidance for the rangeland managers to reduce the occurrence of the overgrazing.

Temporal evolution characteristics and prediction methods of spatial correlation function shape of rough soil surfaces

Soil surface roughness is an important boundary parameter that affects land-atmosphere interactions. Previous studies have primarily characterized the temporal-spatial variability of surface roughness parameters based on root mean square height (rmsh) and correlation length (cl), but ignored the change in correlation function shape of rough soil surfaces. In this study, three rough surfaces with different initial roughness (Rough, Medium, and Small) were designed and the temporal changes in their surface heights were observed from 29 April 2015 to 25 August 2015. Through analyzing the temporal variation of the correlation function shape and its prediction methods, the following conclusions were achieved: 1) the fractal correlation function was more consistent with the experimental data than the exponential and Gaussian function; 2) the α value of fractal correlation function represents the correlation function shape, and it changed from 1 (exponential surface) to 2 (Gaussian surface) from the beginning to the end of the experiment; 3) the α prediction method based on Day of Year (DoY) and cumulative rainfall (CR) was better than the rmsh-based method, due to their feasibility for various initial roughness conditions. In general, this study confirmed that correlation function shape of rough soil surfaces is time-dynamic, and we proposed a method for quantifying its temporal dynamics using DoY or CR with the R2 of 0.71 and 0.74 respectively. The results of this study will help to understand the time-varying characteristics of surface roughness parameters and support future research on agricultural remote sensing and wind erosion.

Temporal Link Prediction: A Survey

The evolutionary behavior of temporal networks has gained the attention of researchers with its ubiquitous applications in a variety of real-world scenarios. Learning evolutionary behavior of networks is directly related to link prediction problem, as the addition or removal of new links or edges over time leads to the network evolution. With the rise of large-scale temporal networks such as social networks, temporal link prediction has become an interesting field of study. In this work, we provide a detailed survey of various researches carried out in the direction of temporal link prediction. We build a taxonomy of temporal link prediction methods based on various approaches used and discuss the works which come under each category. Further, we present the challenges and directions for future works.

A reliable traffic prediction approach for bike‐sharing system by exploiting rich information with temporal link prediction strategy

AbstractBike‐sharing systems have been widely used in major cities across the world. As bike borrowing and return at different stations in different periods are not balanced, the bikes in a bike‐sharing system need to be redistributed frequently to rebalance the system. Therefore, traffic flow forecasting of the bike‐sharing system is an important issue, as this is conducive to achieving rebalancing of the bike system. In this article, we present a new traffic flow prediction approach based on the temporal links in dynamic traffic flow networks. A station clustering algorithm is first introduced to cluster stations into groups. A temporal link prediction method based on the dynamic traffic flow network method (STW+M) is then proposed to predict the traffic flow between stations. In our method, the non‐negative tensor decomposition and time‐series analysis model capture the rich information (temporal variabilities, spatial characteristics, and weather information) of the across‐clusters transition. Then, a temporal link prediction strategy is used to forecast potential links and weights in the traffic flow network by investigating both the network structure and the results of tensor computations. In order to assess the methods proposed in this article, we have used the data of bike‐sharing systems in New York and Washington, DC to conduct bike traffic prediction and the experimental results have shown that our method produces the lowest root mean square error (RMSE) and mean square error (MSE). Compared to four prediction methods from the literature, our RMSE and MSE of the two datasets have been lowered by an average of 2.55 (Washington, DC) and 2.41 (New York) and 3.35 (Washington, DC) and 2.96 (New York), respectively. The results show that the proposed approach improves predictions of traffic flow.