One-dimensional Time Series Research Articles

A novel grey spatial extension relational model and its application to identify the drivers for ambient air quality in Shandong Province, China

The ambient air quality is a complex dynamical system that is shocked by a number of subsystems, such as government policies, industry regulation adjustment and internationalization. To identify the drivers for ambient air quality, a grey spatial extension relational analysis model is proposed. Firstly, a spatial extension method for one-dimensional time series of complex systems is introduced, and the two key parameters are obtained based on the grey similarity and proximity relational analysis models. Secondly, grey relational coefficient is calculated by the difference of the three-dimensional vector, and a grey spatial extension relational analysis model is presented. Furthermore, the properties of the proposed model were investigated. Finally, the model is used to identify the drivers of the ambient air quality in eastern coastal Shandong Province, China. Results suggest that the drivers of the ambient air quality vary among cities, but with some common ones. Therefore, this paper provides an important reference for the improvement of ambient air quality.

AC-LSTM: Anomaly State Perception of Infrared Point Targets Based on CNN+LSTM

Anomaly perception of infrared point targets has high application value in many fields, such as maritime surveillance, airspace surveillance, and early warning systems. This kind of abnormality includes the explosion of the target, the separation between stages, the disintegration caused by the abnormal strike, etc. By extracting the radiation characteristics of continuous frame targets, it is possible to analyze and warn the target state in time. Most anomaly detection methods adopt traditional outlier detection, which has the problems of poor accuracy and a high false alarm rate. Driven by data, this paper proposes a new network structure, called AC-LSTM, which combines Convolutional Neural Networks (CNN) with Long Short-Term Memory (LSTM), and embeds the Periodic Time Series Data Attention module (PTSA). The network can better extract the spatial and temporal characteristics of one-dimensional time series data, and the PTSA module can consider the periodic characteristics of the target in the process of continuous movement, and focus on abnormal data. In addition, this paper also proposes a new time series data enhancement method, which slices and re-amplifies the long time series data. This method significantly improves the accuracy of anomaly detection. Through a large number of experiments, AC-LSTM has achieved higher scores on our collected datasets than other methods.

A method of online anomaly perception and failure prediction for high-speed automatic train protection system

Automatic train protection (ATP) system is the key to ensure the safe operation of high-speed trains. However, the existing operation and maintenance mode for ATP systems cannot diagnose fault in time. In order to improve the protection capability of trains, this paper proposes an online anomaly perception and failure prediction method. First, with real-time operating data, an anomaly perception model based on long short-term memory network is established, where unstructured data are parsed into structured log keys and parameter vectors. It is trained with sequence matrices and its learning performance under different parameters is tested to find the optimal model. Experimental results show that the classification accuracy is 0.981, which is better than the existing methods. Then, with historical data, a failure prediction model based on time series is established, where one-dimensional time series of failure rate are reconstructed to high-dimensional space. The support vector regression method is used to fit the complex functional relationship between phase point and predicted point. And different algorithms are taken to find the optimal parameters. The results show that the model has the strongest generalization ability with the accuracy of 0.987. Finally, the intelligent operation and maintenance data service platform is designed.

Impacts of the Covid-19 crisis on unemployment in Slovakia: a statistically created counterfactual approach using the time series analysis

Research background: The current health crisis, which began to take shape in mid-March 2020 due to the massive spread of the coronavirus SARS-CoV-2, has shaped the face of the Slovak labour market the most. Slovakia, similarly, to other countries in the world, has operated with several non-pharmaceutical anti-pandemic measures taken by the government to reduce the spread of the virus or reverse adverse developments. On the other hand, these measures have frozen production, and business activities of companies annulled not only the number of physical meetings with business collaborators but also business contracts themselves. Consequently, the demand for new labour has decreased, and in many cases, redundancies have occurred. Purpose of the article: The study aims to develop a comprehensive and detailed analysis of unemployment and its development in Slovakia, during the 1st and 2nd wave of the Covid-19 pandemic, determine the extent to which the anti-pandemic measures have affected the Slovak labour market, and identify the most affected groups of unemployed. Methods: To analyze the impact of the anti-pandemic measures on the unemployment situation in Slovakia, we applied the approach where the statistically created hypothetical state of the absence of the pandemic was compared with the real state as the consequence of the pandemic. The hypothetical state was modelled by the autoregressive 1st order or one-dimensional time series model with a linear trend and seasonability. Findings & value added: The results of the analysis identify the groups of the population, mostly affected by the Covid-19 pandemic together with the quantification of the impact on unemployment. The findings of quantification of the impact of the Covid-19 crisis on the Slovak labour market can be further used in creating targeted measures aimed at supporting unemployed individuals or at the sustainability of already existing jobs and in other governmental economic and social decisions.

FORECASTING PASSENGER TRAFFIC BASED ON STRUCTURE IDENTIFICATION OF NON-STATIONARY TIME SERIES

The article proposes to consider the change in the volume of passenger traffic as non-stationary time series which is characterized by the presence of transients moments that lead to a significant change in the overall trend over time. Studies of changes in demand for passenger traffic in suburban service on bus routes of Cherkasy region in 2019–2021 showed that in the period up to 2020, the daily amount of presented demand on weekdays ranged from 4.1 thousand to 6.4 thousand passengers, and on the second half of 2020, the volume of demand decreased due to applying quarantine restrictions in 1.9–2.1 times (up to 2.1 thousand to 4.1 thousand passengers). Based on the analysis of fluctuations in the actual volume of traffic, the structural and logical inter-component relationship of the formation of demand for transportation was identified, which consists of the allocation of the main groups of transport classes forming the volume of passengers. By adapting the methodological provisions of singular spectral analysis, the main stages of the procedure for identifying the elemental structure of the non-stationary time series of changes in the volume of transportation have been formalized. Based on the analysis of the preconditions for the emergence of transients information on demand for passenger traffic allocated structure elements of the time series of the actual volume of passenger traffic that is based on the features and factors of its occurrence. Implementation of the procedure of using the singular spectral method analysis in the study of changes in the volume of passenger traffic is proposed to do by converting a one-dimensional time series into a matrix of elements that determine the trend, harmonic and random components demand. Based on the analysis of the materials of the passenger survey, it was established that the trend component of the change in demand in the Cherkasy – Smila connection contains two types of labor movements: permanent nature of formation and temporary. For the studied combination during 2019–2021, the share of worker transfers with a permanent character increased from 46.5% to 53.1%, with temporary character – decreased by 30.2% to 20.3%, and the harmonic component - increased from 11.5% to 15.3%, and the random component - remained stable within 11.9% to 11.2%. Based on the synthesis of predicted values of time components series for the connection Cherkasy – Smila found that in terms of implementation positive development scenario, there is a tendency to gradual short-term growth in traffic by 3.5%.

Robust Dynamic Hand Gesture Recognition Based on Millimeter Wave Radar Using Atten-TsNN

The millimeter-wave radar can sense the subtle movement of the hand. However, the traditional hand gesture recognition methods do not work well in dynamic interference scenarios. To address this issue, a robust hand gesture recognition method is proposed based on the self-attention time-series neural networks. Firstly, the original radar echo is constructed in terms of frame, sequence, and channel at the input terminal of the network. A one-dimensional time-series neural network is built, and the time-distributed layer is used as the wrapper to extract the feature from each frame sequence independently. Then the self-attention mechanism is employed to assign the adequate weights to the sequence of frames entered in parallel to obtain the inter-frame correlation and suppress the random interference. Finally, the Global AvgPooling layer is used to reduce the number of channels, and the fully connected layer outputs the label of the gesture. The experimental results show that the proposed method can achieve a high recognition rate in the presence of random dynamic interference.

Fault diagnosis of rolling bearing based on singular spectrum analysis and wide convolution kernel neural network

Research on bearing fault diagnosis generally uses a single signal processing method to extract features to a certain extent and then empirically judges the fault type at this stage. However, this approach has a poor fault feature extraction effect when the noise interference signal is large and subjective human errors. The diagnosis effect is determined by external interference and has no practical value. This paper proposes a rolling bearing fault diagnosis method based on singular spectrum analysis and a wide convolution kernel neural network, which can effectively extract the fault features of the rolling bearing crack fault signal with a strong noise interference and realize the efficient diagnosis of this kind of fault. Gaussian white noise is added to the standard bearing fault signal to construct the vibration signal with noise interference. Then, the noisy one-dimensional time series signal is preprocessed, including the data division into the training and test sets and the different kinds of numbering processing. Singular spectrum analysis is performed on the preprocessed data. Then, the denoised training set is used as input in a deep convolution neural network with a wide convolution kernel for feature extraction and model training. The trained diagnosis model is used for the fault prediction of the next test set, and the relevant diagnosis results are output. The test results show that this method can ensure the overall accuracy of more than 90% under the background of high noise, and the diagnosis rate of the model under various working conditions is stably maintained at more than 93%, without the collapse of diagnosis stability under the sudden change of noise. The advantages of model diagnosis efficiency and structural improvement fit are prominent.

The prediction model for haze pollution based on stacking framework and feature extraction of time series images

In this paper, we propose a new model called “image-feature-stacking prediction model” to study the prediction problem of univariate time series data. Its main idea is to convert univariate time series data into corresponding images, and then use the optimized Inception-v1 network to extract hidden features from the images as input variables, based on these features, a two-layer stacking ensemble learning framework is constructed to output the final predicted values. The main contribution of the newly proposed model is to convert one-dimensional time series data into two-dimensional images, and automatically extract features from images. This method can truly mine the intrinsic relationship between the data instead of simply relying on descriptive statistical features to replace the original time series, thereby improving the prediction performance of the model. We use the new prediction model to predict daily PM2.5 concentration, for one-step prediction, the results show that compared with the other three time series prediction models, the proposed prediction model reduces the mean absolute percentage error and mean absolute scaled error to 19.204% and 1.242, respectively, which is 76.607% and 77.004% lower than the maximum value of mean absolute percentage error and mean absolute scaled error of four prediction models. We also make two-step and three-step predictions, and the newly proposed model also shows encouraging performance.

Ferromagnetic resonance over-voltage identification method based on Gram angle field

In neutral point non-grounding systems with 35 kV and below, ferromagnetic resonance over-voltage faults often occurs, which brings safety hazards to power systems. Ferromagnetic resonance occurs for a short time and is similar to other faults in the power system and is not easy to be detected by direct observation. Deep learning has obvious advantages in image recognition and classification, but it is difficult to deal with over-voltage one-dimensional time series monitored in power system, and a deep learning recognition model based on Gram angle field and improving Convolutional Neural Network (CNN) structure is proposed, the voltage time series is converted into two-dimensional image, and the voltage classification task is performed by extracting image features and entering them into two channels of CNN-SPP (Spatial Pyramid Pooling). The two-channel structure solves the problem of high-dimensional time series learning, SPP solves the problem that CNN requires the input over-voltage conversion image must be the same size, and optimizes the network structure and parameters to obtain the optimal identification model. Finally, the experimental comparison shows that the ferromagnetic resonant over-voltage voltage recognition based on Gram angle field and improving CNN has higher recognition accuracy than other time domain and frequency domain characteristic algorithms. The high frequency, division frequency and fundamental frequency ferroresonance identification accurate rate based on the proposed method is 94.2%, 93.8% and 90.5% respectively.

Evaluation of Turbulent Jet Characteristic Scales Using Joint Statistical Moments and an Adaptive Time-Frequency Analysis

This paper presents an analysis of turbulence characteristic scales and eddy convection velocity of jet flows computed using joint statistical moments, digital filters, and a modified version of the empirical mode decomposition (EMD). The ongoing aim of this study is to develop semi-empirical space-time cross-correlation models based on stationary statistics and jet physical lengths. Multivariate statistics are used to correlate jet properties to one-dimensional time series. The data available to this study were recorded from single-point and two-point hot-wire anemometry experiments carried out for a range of jet Mach numbers (0.2≤M≤0.8). Firstly, the jet eddy convection velocity, turbulence length, and time scales are computed using space-time cross-correlation functions. Isotropic flow and frozen turbulence hypothesis are then used to estimate the joint moments from single-point statistics in the fully developed turbulence region. An EMD-based decomposition method is presented and assessed in both the Gaussian and non-Gaussian signal regions. It is demonstrated that the artificially filtered signal reconstructs the physical properties of single and multi-point jet statistics. The relationship between central moments and joint moments presented here focuses on the region of high turbulence levels, which generates the vast majority of jet mixing noise produced by turbofan engines. Further analysis is required to extend this investigation to intermittent zones and other jet noise sources, such as jet-surface installation noise.

Statistical inference of one-dimensional persistent nonlinear time series and application to predictions

We introduce a method for reconstructing macroscopic models of one-dimensional stochastic processes with long-range correlations from sparsely sampled time series by combining fractional calculus and discrete-time Langevin equations. The method is illustrated for the ARFIMA(1,d,0) process and a nonlinear auto-regressive toy model with multiplicative noise. We reconstruct a model for daily mean temperature data recorded at Potsdam (Germany) and use it to predict the first frost date by computing the mean first passage time of the reconstructed process and the zero degree Celsius temperature line, illustrating the potential of long-memory models for predictions in the subseasonal-to-seasonal range.

Analysis of Symmetrical and Unsymmetrical Faults Using the EEMD and Scale-Dependent Intrinsic Entropies

The rapid and accurate diagnosis of power grid faults plays a vital role in speeding up the process of accident handling and system recovery and ensuring the safe operation of the power system. This paper proposes to apply the ensemble empirical mode decomposition (EEMD) method and scale-related intrinsic entropy to diagnose the type of fault for the transmission line. First, the electrical data collected by the power system is decomposed by using the EEMD method. Then by eliminating some intrinsic mode functions, the signal is reconstructed by inspecting the correlation coefficient. Finally, the complexity of the reconstructed signal is calculated by using the scale-dependent intrinsic entropy. Since the scale-dependent intrinsic entropy reflects the complexity of one-dimensional time series, it is susceptible to signal changes. The complexity is helpful in the power system for fault signal analysis. The results show the combined method’s effectiveness and practicability through failure analysis using the IEEE 14-bus system as the simulation model.

A short-term flood prediction based on spatial deep learning network: A case study for Xi County, China

Floods cause substantial damage across the world every year. Accurate and timely prediction of floods can significantly minimize the loss of life and property. Recently, numerous machine learning models have been used for flood prediction, showing that their performance is preferable to traditional statistical models. However, the existing models neglect the spatial features of floods, which drive flood generation and concentration. In this paper, the area of interest is divided into grids based on longitude and latitude, and the rainfall and discharge collected by stations are combined into tensors according to station coordinates. Different from one-dimensional time series, our input feature is a two-dimensional time series with spatial information. Hence, combining a Convolutional Neural Network (CNN) with a Long Short Term Memory Network (LSTM), we propose the convolution LSTM (ConvLSTM) to extract spatiotemporal features of hydrological information. The methodology is demonstrated using the hydrological data collected at the Xi County stations, located on the Huai River in Henan Province, China. Numerical results indicate that the relative error of arrival time is within 30%, and the relative error of peak discharge is within 20%, satisfying the 2005 Chinese Water Resource Standard on flood prediction permit error. The experiments also show that the ConvLSTM outperforms the recent models in terms of flood arrival time and peak discharge, thereby proving a promising alternative.

Tangle: A metric for quantifying complexity and erratic behavior in short time series.

Temporal complexity refers to qualities of a time series that are emergent, erratic, or not easily described by linear processes. Quantifying temporal complexity within a system is key to understanding the time based dynamics of said system. However, many current methods of complexity quantification are not widely used in psychological research because of their technical difficulty, computational intensity, or large number of required data samples. These requirements impede the study of complexity in many areas of psychological science. A method is presented, tangle, which overcomes these difficulties and allows for complexity quantification in relatively short time series, such as those typically obtained from psychological studies. Tangle is a measure of how dissimilar a given process is from simple periodic motion. Tangle relies on the use of a three-dimensional time delay embedding of a one-dimensional time series. This embedding is then iteratively scaled and premultiplied by a modified upshift matrix until a convergence criterion is reached. The efficacy of tangle is shown on five mathematical time series and using emotional stability, anxiety time series data obtained from 65 socially anxious participants over a 5-week period, and positive affect time series derived from a single participant who experienced a major depression episode during measurement. Simulation results show tangle is able to distinguish between different complex temporal systems in time series with as few as 50 samples. Tangle shows promise as a reliable quantification of irregular behavior of a time series. Unlike many other complexity quantification metrics, tangle is technically simple to implement and is able to uncover meaningful information about time series derived from psychological research studies. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

General Tensor Least-Mean-Squares Filter for Multi-Channel Multi-Relational Signals

Least-mean-squares (LMS) algorithms constitute a prevalent approach to implement the linear adaptive filters whose coefficients can be updated sample by sample so as to track time-varying dynamics. As the memory and computational complexities required for the realization of LMS filters are very low, they have been widely adopted in many real-time signal processing applications. The input of any conventional LMS filter has to be a sequence of scalar samples (one-dimensional time series), whereas such assumption is too restrictive nowadays for multi-channel (high-dimensional) signals and multi-relational data in the rise of a big-data era. It is crucial to deal with high-dimensional data-arrays, a.k.a. tensors, to manifest the variety and complex interrelations of data. Owing to lack of a sufficient mathematical framework to govern relevant tensor operations, the general tensor LMS filter, whose input is allowed to be an arbitrary tensor, has never been established for realization to the best of our knowledge. In this work, we will dedicate a new mathematical framework for tensors to establish the general tensor least-mean-squares (TLMS) filter theory and propose two novel TLMS algorithms with update rules based on stochastic gradient-descent and Newton's methods, respectively. Furthermore, as we establish the tensor calculus theory, the performance evaluation on convergence-rate and misadjustment for our proposed TLMS filters can be conducted. Finally, the memory and computational complexities of the new TLMS algorithms are also studied in this paper.

ПОДХОД К ОЦЕНКЕ КАЧЕСТВА ОБНАРУЖЕНИЯ АНОМАЛИЙ В ТЕХНОЛОГИЧЕСКИХ СИГНАЛАХ

A technological signal is a one-dimensional time series, which is an ordered sequence of discrete-time data. Due to the presence of time dimension, methods for detecting anomalies in time series should take into account time correlations and other time-related features. As a rule, the inaccuracy matrix and metrics derived from it, such as accuracy, completeness, F-measure, etc., are used to assess the quality of the anomaly detection method. These metrics, however, do not take into account the presence of time dimension. The metric, composed of the F-measure and distance-based metrics, allows you to take into account the moment of the beginning of the anomaly, the balance between errors of the first and second kind, the presence of an anomalous site. The paper proposes an approach to the construction of a quality metric for detecting anomalies in technological signals that comprehensively components these characteristics. Due to this arrangement, the metric evaluates the time dimension of the data, allows for a more adequate assessment of the presence of point anomalies and the occurrence of abnormal areas in technological signals, taking into account their characteristics, separating them from regular (normal, typical) signals. The integral assessment takes into account not only various features of the data, but can also be configured taking into account the specifics of a specific task. The paper presents the results of testing the proposed quality metric, which showed the effectiveness of the proposed approach to assessing the quality of anomaly detection in technological signals. The superiority of the proposed metric in the considered situations was on average more than 10%. At the same time, an additional key advantage of the metric over its analogues is the ability to configure it taking into account the specifics of the data and the models used for detecting anomalies.

Parameter Prediction of Marine Seawater Cooling System Based on Chaos-Elman Combined Model

To improve the accuracy of short-term prediction of marine seawater cooling system parameters and reduce the accuracy loss caused by chaotic data, a combined prediction model combining chaos theory and Elman feedback neural network was proposed in this paper. Firstly, the C-C algorithm and G-P algorithm were used to calculate the two reconstruction parameters respectively, and according to the properties of the maximum Lyapunov exponent, it could be determined that the original one-dimensional time series had chaotic characteristics. Secondly, phase space reconstruction technique was used to map the one-dimensional seawater outlet temperature time series of intercooler to the high-dimensional space to find its chaotic phase trajectory, and the number of adjacent points was determined by HQ (Hannan Quinn) information criterion. Finally, the Chaos-Elman combined prediction model was used to train and test 300 sets of seawater outlet temperature data of intercooler, and the accuracy of prediction was compared with that of single structure model. The experimental results were as follows: the average prediction accuracy based on chaos theory was 95.5%, the prediction accuracy of Elman neural network was 95.1%, and the prediction accuracy of Chaos-Elman combined model could reach 98.7%. In this study, the Chaos-Elman combined model has better prediction accuracy than the single model. It can be concluded that the Chaos-Elman combined model can be used for the short-term prediction of seawater cooling system parameters, and the prediction accuracy and reliability are high.

МЕТОДЫ ИССЛЕДОВАНИЯ СТОИМОСТИ ПРОДУКТОВОЙ КОРЗИНЫ ПО ОРЕНБУРГСКОЙ ОБЛАСТИ

The cost of a minimum set of food takes an essential place in the structure of the subsistence minimum, and largely determines the conditions of existence of the poorest segments of the population. In modern conditions of rising prices for vital foodstuffs, it is essential to analyze the cost of these goods in order to timely prevent the negative consequences of such an increase. The article is devoted to the assessment of changes in the actual average and minimum physiologically necessary living standards of citizens of the Russian Federation through a comparison of inflation indices calculated on the basis of consumer baskets and the index of changes in the value of average wages. Special attention is paid to the comparative analysis based on one-dimensional time series of the dynamics of the cost of five food baskets, the transition to comparable prices is reasoned. It is noted that food products can be divided into two groups. Prices for some grow monotonously, without the influence of the time of year, which is more typical for industrial goods. The second subset is products with pronounced seasonality, primarily vegetables, whose prices fall in the second half of summer and autumn, and then begin to increase. The presence of this component leads to the fact that the growth of basket values practically stops in summer, and the fastest is in winter. The contributions of individual food products to the cost of consumer baskets have been studied. Attention is paid to the difference between normative (i.e., given a priori) baskets and the baskets obtained as a result of the analysis of real consumption. The article reflects not only theoretical aspects of the analysis of the standard of living of the population and the cost of the grocery basket, but also practical results. The study of changes in the structure of the cost of each of the considered food baskets caused by changes in prices for the corresponding goods is carried out, the influence of various socio-economic indicators is not taken into account.

Online Transient Frequency Safety Prediction Machine of Power System Based on Time-Feature Attention Module

Facing the frequency safety problem of renewable energy power systems, this paper proposes a Frequency Safety Prediction Machine (FSPM) based on a one-dimensional convolutional neural network (1D-CNN) and Time-Feature Attention Module (TFAM). To ensure the rapidity of online frequency safety prediction (FSP), the one-dimensional time series data (1D-TSD) at multiple moments are standardized and divided into multiple time channels and transmitted to FSPM at the same time, which speeds up the speed of data flow. The TFAM in FSPM realizes the function of focusing on key moments and key features. Make FSPM pay more attention to the system information at certain critical moments and certain key feature information that can fully represent the fault, thereby improving the prediction accuracy. In addition, FSPM achieves dual prediction of Frequency Danger Level (FDL) and Time Safety Margin (TSM) at the same time. Finally, the test results on the IEEE 39 bus system and the Illinois power grid both containing renewable energy show that the FSPM not only has good performance in both FDL and TSM prediction but also has low model complexity, good noise resistance, and robustness.

Artificial Intelligence-Based Echocardiographic Left Atrial Volume Measurement with Pulmonary Vein Comparison.

This paper combines echocardiographic signal processing and artificial intelligence technology to propose a deep neural network model adapted to echocardiographic signals to achieve left atrial volume measurement and automatic assessment of pulmonary veins efficiently and quickly. Based on the echocardiographic signal generation mechanism and detection method, an experimental scheme for the echocardiographic signal acquisition was designed. The echocardiographic signal data of healthy subjects were measured in four different experimental states, and a database of left atrial volume measurements and pulmonary veins was constructed. Combining the correspondence between ECG signals and echocardiographic signals in the time domain, a series of preprocessing such as denoising, feature point localization, and segmentation of the cardiac cycle was realized by wavelet transform and threshold method to complete the data collection. This paper proposes a comparative model based on artificial intelligence, adapts to the characteristics of one-dimensional time-series echocardiographic signals, automatically extracts the deep features of echocardiographic signals, effectively reduces the subjective influence of manual feature selection, and realizes the automatic classification and evaluation of human left atrial volume measurement and pulmonary veins under different states. The experimental results show that the proposed BP neural network model has good adaptability and classification performance in the tasks of LV volume measurement and pulmonary vein automatic classification evaluation and achieves an average test accuracy of over 96.58%. The average root-mean-square error percentage of signal compression is only 0.65% by extracting the coding features of the original echocardiographic signal through the convolutional autoencoder, which completes the signal compression with low loss. Comparing the training time and classification accuracy of the LSTM network with the original signal and encoded features, the experimental results show that the AI model can greatly reduce the model training time cost and achieve an average accuracy of 97.97% in the test set and increase the real-time performance of the left atrial volume measurement and pulmonary vein evaluation as well as the security of the data transmission process, which is very important for the comparison of left atrial volume measurement and pulmonary vein. It is of great practical importance to compare left atrial volume measurements with pulmonary veins.