Airport security check passenger flow prediction method and system, electronic equipment and storage medium

Abstract

The invention provides an airport security check passenger flow volume prediction method and system, electronic equipment and a storage medium, and is applied to the technical field of airport security check. Target historical airport security check queuing data before target time is acquired; fusing the queue information of each target historical queue to obtain fused queue information of a target fused historical queue; performing empirical mode decomposition processing on fusion queue information of the target fusion historical queue to obtain a plurality of target data sequences and target residual components; determining a prediction model matched with the target data sequence, and performing airport security check passenger flow prediction according to the target data sequence or the target data sequence and the target residual component through the prediction model matched with the target data sequence to obtain a target prediction result; and determining the airport security check passenger flow volume corresponding to the target time according to each target prediction result, thereby being capable of adapting to unstable data, meeting actual requirements, and improving the accuracy of predicting the airport security check passenger flow volume at the same time.

Description

Technical Field

[0001] This invention relates to the field of airport security inspection technology, and more specifically, to a method, system, electronic device, and storage medium for predicting passenger flow at airport security checkpoints. Background Technology

[0002] With the rapid development of air transport, airport passenger traffic is constantly increasing. As an important part of air transport, the efficiency of airport security checks directly affects the travel experience of passengers and the operational efficiency of the airport. Therefore, how to accurately predict the passenger flow at airport security checkpoints is an urgent problem to be solved in order to optimize the allocation of airport security checkpoint resources, improve security check efficiency, reduce passenger waiting time, and thus improve the travel experience of passengers and the operational efficiency of the airport.

[0003] Current methods for predicting airport security passenger flow mainly rely on manual prediction based on historical airport security passenger flow. This method is not only inaccurate but also poorly adaptable to unstable data, making it difficult to meet actual needs. Summary of the Invention

[0004] In view of this, this application provides a method, system, electronic device and storage medium for predicting airport security passenger flow, in order to adapt to unstable data, meet actual needs and improve the accuracy of predicting airport security passenger flow.

[0005] The first aspect of this application provides a method for predicting passenger flow at airport security checkpoints, the method comprising:

[0006] Determine the target time for airport passenger traffic forecasting and obtain the target historical airport security queue data prior to the target time; wherein, the target historical airport security queue data includes queue information of multiple target historical queues;

[0007] The queue information of each target historical queue is merged to obtain the merged queue information of the target merged historical queue;

[0008] Empirical mode decomposition is performed on the fusion queue information of the target fusion history queue to obtain multiple target data sequences and target residual components;

[0009] For each target data sequence, a prediction model matching the target data sequence is determined, and airport security passenger flow is predicted based on the target data sequence or the target data sequence and the target residual component using the prediction model matching the target data sequence to obtain the target prediction result; wherein, each prediction model is obtained by training the prediction model to be trained using each data sequence obtained by processing historical airport security queue data and training residual component or each data sequence.

[0010] The airport security passenger flow corresponding to the target time is determined based on the prediction results of each target.

[0011] Optionally, the queue information of the target history queue includes multiple consecutive target history times and the number of people queuing for each target history time;

[0012] The step of fusing the queue information of each of the target historical queues to obtain the fused queue information of the target fused historical queue includes:

[0013] The number of people queuing in each of the target historical queues belonging to the same target historical time is merged to obtain the merged queue information of the initial merged historical queue;

[0014] The fusion queue information of the initial fusion history queue is preprocessed to obtain the fusion queue information of the target fusion history queue.

[0015] Optionally, the step of performing empirical mode decomposition on the fusion queue information of the target fusion history queue to obtain multiple target data sequences and target residual components includes:

[0016] The original target signal is generated based on the fusion queue information of the target fusion history queue;

[0017] Extract all local extrema from the original target signal;

[0018] An upper envelope is generated based on the local maxima among the local extrema, and a lower envelope is generated based on the local minima among the local extrema.

[0019] A mean envelope is generated based on the upper envelope and the lower envelope, and a target intermediate signal is generated based on the target original signal and the mean envelope.

[0020] Based on the target intermediate signal and the target mean envelope, the next target intermediate signal is determined, and a target data sequence and its sequence number are generated based on the next target intermediate signal; wherein, the target mean envelope is determined based on the target intermediate signal;

[0021] The residual component is determined based on the original target signal and the intermediate signal of the next target;

[0022] Determine whether the target intermediate signal satisfies the data sequence conditions;

[0023] If the target intermediate signal satisfies the data sequence condition, the residual component is updated to the target original signal, and the process returns to extracting all local extrema from the target original signal.

[0024] If the target intermediate signal does not meet the data sequence conditions, the target residual component is determined based on the residual component, and the empirical mode decomposition process of the fusion queue information of the target fusion history queue is completed.

[0025] Optionally, for each target data sequence, determining a prediction model matching the target data sequence, and using the prediction model matching the target data sequence to predict airport security passenger flow based on the target data sequence or the target data sequence and the target residual component, to obtain the target prediction result, includes:

[0026] For each target data sequence, the prediction model with the same index as the target data sequence is determined as the prediction model that matches the target data sequence.

[0027] Determine whether the training data used by the prediction model during the training process is a training data sequence, or whether it is a training data sequence and training residual components;

[0028] If the training data used by the prediction model during the training process is a training data sequence, the target data sequence is input into the prediction model that matches it, so that the prediction model that matches the target data sequence makes a prediction based on the target data sequence and obtains the target prediction result corresponding to the target data sequence.

[0029] If the training data used by the prediction model during training is a training data sequence and a training residual component, the target data sequence and the target residual component are input into the prediction model that matches it, so that the prediction model that matches the target data sequence makes a prediction based on the target data sequence and the target residual component, and obtains the target prediction result corresponding to the target data sequence.

[0030] Optionally, the process of processing historical airport security queuing data to obtain various data sequences and residual components, or various data sequences, is used to train the prediction model to be trained, resulting in multiple prediction models, including:

[0031] Collect historical airport security check queue data; wherein, the historical airport security check queue data includes airport security check passenger flow sequences, and the airport security check passenger flow sequences include queue information of multiple historical queues;

[0032] Each of the historical queues is divided into a training set and a test set; wherein the training set includes multiple training queues and the test set includes multiple test queues.

[0033] The queue information of each training queue is merged to obtain the merged queue information of the training queue, and the queue information of each test queue is merged to obtain the merged queue information of the test queue.

[0034] The fusion queue information of the fusion training queue is processed by empirical mode decomposition to obtain multiple training data sequences and training residual components arranged in sequence; wherein, each training data sequence has a corresponding sequence number;

[0035] The fusion queue information of the fusion test queue is processed by empirical mode decomposition to obtain multiple test data sequences and test residual components arranged in sequence; wherein, each test data sequence has a corresponding sequence number;

[0036] For each index, the prediction model to be trained is trained using the training residual components, the training data sequence with the same index, and the test data sequence to obtain the prediction model under that index.

[0037] Optionally, the step of training the prediction model to be trained using the training residual components, the training data sequence with the same index, and the test data sequence to obtain the prediction model under that index includes:

[0038] The training data sequence is input into the prediction model to be trained, so that the prediction model to be trained can be trained using the training data sequence to obtain an initial prediction model.

[0039] The initial prediction model is validated using a test data sequence with the same sequence number as the training data sequence to obtain the goodness of fit of the initial prediction model;

[0040] If the goodness of fit of the initial prediction model is greater than the preset goodness of fit, the initial prediction model is determined to be the prediction model under the sequence number of the training data sequence; wherein, the training data used during the training of the prediction model is the training data sequence.

[0041] If the goodness of fit of the initial prediction model is not greater than the preset goodness of fit, the training data sequence and the training residual components are input into the initial prediction model, so that the initial prediction model is trained using the training data sequence and the training residual components to obtain the prediction model under the sequence number of the training data sequence; wherein, the training data used when the prediction model is trained are the training data sequence and the training residual components.

[0042] A second aspect of this application provides an airport security check passenger flow prediction system, the system comprising:

[0043] The data acquisition module is used to determine the target time for airport passenger flow prediction and to acquire the target historical airport security check queue data before the target time; wherein, the target historical airport security check queue data includes queue information of multiple target historical queues;

[0044] The data preprocessing module is used to fuse the queue information of each target historical queue to obtain the fused queue information of the target fused historical queue;

[0045] The EMD decomposition module is used to perform empirical mode decomposition processing on the fusion queue information of the target fusion history queue to obtain multiple target data sequences and target residual components.

[0046] The prediction module is used to determine a prediction model matching the target data sequence for each target data sequence, and to predict airport security passenger flow based on the target data sequence or the target data sequence and the target residual component using the prediction model matching the target data sequence, thereby obtaining a target prediction result; wherein, each prediction model is obtained by the training module using each data sequence obtained by processing historical airport security queue data and training residual components or each data sequence to be trained prediction model;

[0047] The result reconstruction module is used to determine the airport security passenger flow corresponding to the target time based on the prediction results of each target.

[0048] Optionally, the queue information of the target historical queue includes multiple consecutive target historical times and the number of people queuing for each target historical time. The data preprocessing module includes:

[0049] The fusion unit is used to merge the number of people queuing in the same target historical time in each of the target historical queues to obtain the fusion queue information of the initial fusion historical queue.

[0050] The preprocessing unit is used to preprocess the fusion queue information of the initial fusion history queue to obtain the fusion queue information of the target fusion history queue.

[0051] A third aspect of this application provides an electronic device, comprising: a processor and a memory, the processor and the memory being connected via a bus; wherein the processor is configured to call and execute a program stored in the memory; and the memory is configured to store the program, the program being configured to implement an airport security passenger flow prediction method as provided in the first aspect of this application.

[0052] The fourth aspect of this application provides a computer-readable storage medium storing computer-executable instructions for performing an airport security passenger flow prediction method as provided in the first aspect of this application.

[0053] This application provides a method, system, electronic device, and storage medium for predicting airport security passenger flow. The method involves determining the target time for airport passenger flow prediction and acquiring historical airport security queue data prior to the target time. The historical airport security queue data includes queue information from multiple historical queues. The queue information from each historical queue is fused to obtain fused queue information for the fused historical queue. Empirical mode decomposition (EMD) is applied to the fused queue information to obtain multiple target data sequences and target residual components. For each target data sequence, a prediction model matching the target data sequence is determined, and airport security passenger flow is predicted based on the target data sequence or the target data sequence and the target residual components using the prediction model matching the target data sequence, resulting in a target prediction result. Each prediction model is trained using the data sequences obtained by processing historical airport security queue data and training residual components or the data sequences to be trained. The airport security passenger flow corresponding to the target time is determined based on the target prediction results. Therefore, the technical solution provided in this application can effectively decompose the original non-stationary time series data by performing empirical mode decomposition on the target historical airport security check data, remove noise from the data, and extract multiple target data sequences and target residual components. Then, each target data sequence or each target data sequence and target residual component is input into a matching prediction model for prediction. By combining the time series prediction capability of the prediction model, the prediction accuracy of complex non-stationary time series data can be effectively improved, that is, the accuracy of predicting airport security check passenger flow can be improved. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0055] Figure 1 A flowchart illustrating an airport security passenger flow prediction method provided in this application embodiment;

[0056] Figure 2 An example diagram illustrating a method for training a prediction model, as provided in an embodiment of this application;

[0057] Figure 3 A schematic diagram of the structure of an airport security passenger flow prediction system provided in this application embodiment;

[0058] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] In this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0061] To better understand the technical terms used in this application, the following explanations are provided:

[0062] Empirical Mode Decomposition (EMD) is an adaptive signal processing method primarily used to decompose complex non-stationary signals into several Intrinsic Mode Functions (IMFs) and a residual component. In other words, it is a means of stabilizing non-stationary signals by progressively decomposing the fluctuations and trends at different scales in the signal to generate a series of data sequences with different characteristic scales. Each data sequence is called an Intrinsic Mode Function.

[0063] Long Short-Term Memory (LSTM) networks are a special type of recurrent neural network (RNN) architecture specifically designed to handle long- and short-term dependencies in time-series data. LSTMs effectively address the vanishing or exploding gradient problems that traditional RNNs often encounter when processing long-series data by introducing gating mechanisms (including forget gates, input gates, and output gates) to control the flow of information. This allows LSTMs to capture long-term dependencies in time-series data and makes them widely used in time-series forecasting, natural language processing, and other fields.

[0064] Ensemble learning algorithms (eXtreme Gradient Boosting, XGBoost) improve prediction performance by building and combining multiple weak classifiers (usually decision trees).

[0065] As can be seen from the background technology above, the current methods for predicting airport security passenger flow mainly rely on manual prediction based on historical airport security passenger flow. This method not only has poor prediction accuracy but also poor adaptability to unstable data, making it difficult to meet actual needs.

[0066] In existing technologies, airport security passenger flow prediction can be performed using XGBoost, variational autoencoders combined with LSTM, or attention mechanisms. However, XGBoost cannot effectively model long-term temporal dependencies, and while suitable for low-noise data, its ability to handle non-stationary data like airport passenger flow is limited. Variational autoencoders combined with LSTM have high computational complexity and are not suitable for real-time prediction. Attention mechanisms, while suitable for long-sequence prediction, have high requirements for small-scale data and consume significant computational resources.

[0067] Therefore, this application provides a method, system, electronic device, and storage medium for predicting airport security passenger flow. By performing empirical mode decomposition on historical airport security data, the original non-stationary time series data can be effectively decomposed, noise removed, and multiple target data sequences and target residual components extracted. Each target data sequence, or each target data sequence and target residual component, is then input into a matching prediction model for prediction. By combining the time series prediction capabilities of the prediction model, the prediction accuracy for complex non-stationary time series data is effectively improved, thus providing higher accuracy in predicting airport security passenger flow. Therefore, this method is not only applicable to low-noise data, has strong processing capabilities for non-stationary data such as airport passenger flow, low computational complexity, and suitable for real-time prediction, but also has low requirements for small-scale data and low computational resource consumption.

[0068] See Figure 1 The diagram illustrates a flowchart of an airport security passenger flow prediction method provided in an embodiment of this application. The method specifically includes the following steps:

[0069] S101: Determine the target time for the airport passenger flow forecast and obtain the target historical airport security check queue data before the target time; wherein, the target historical airport security check queue data includes queue information of multiple target historical queues.

[0070] In the specific execution step S101, the target time for predicting airport passenger flow can be determined first, so as to obtain the target historical airport queuing data before the target time. Specifically, a corresponding preset time period can be set in advance, and then the historical airport queuing data within the preset time period before the target time can be obtained (for ease of distinction, the historical scenario queuing data before the target time is referred to as the target historical airport queuing data). Among them, the target historical airport queuing data includes the target historical airport passenger flow sequence and its sequence information, which is composed of multiple target historical queues and their queue information.

[0071] It should be noted that the queue information for each target historical queue includes multiple target historical moments ordered in chronological order and the number of people in the queue corresponding to each historical target historical moment.

[0072] It should also be noted that the preset time period can be 0.5 hours, 1 hour, 2 hours, 10 hours, 100 hours, 200 hours, etc., and is not limited in this embodiment of the application.

[0073] S102: Merge the queue information of each target historical queue to obtain the merged queue information of the target merged historical queue.

[0074] In the specific execution of step 102, after obtaining the target historical airport queuing data, the number of people queuing in the queue information of each target historical queue in the target historical airport queuing data can be merged to obtain a target merged historical queue and its merged queue information.

[0075] Optionally, the process of fusing the queue information of each target historical queue to obtain the fused queue information of the target fused historical queue can be as follows: fused the number of people queuing in each target historical queue belonging to the same target historical time to obtain the fused queue information of the initial fused historical queue; preprocessed the fused queue information of the initial fused historical queue to obtain the fused queue information of the target fused historical queue.

[0076] It should be noted that the initial fused historical queue information includes the number of people queuing at each target historical moment, sorted chronologically. By cleaning the initial fused historical queue information to remove null and erroneous values, and then normalizing the cleaned information to scale the number of people queuing at each target historical moment to the [0, 1] interval, the target fused historical queue information is obtained. This improves the accuracy of predicting airport security passenger flow in subsequent airport security passenger flow predictions using the target fused historical queue information.

[0077] S103: Perform empirical mode decomposition on the fusion queue information of the target fusion history queue to obtain multiple target data sequences and target residual components.

[0078] In the specific execution step S103, after obtaining the fusion queue information of the target fusion history queue, the target original signal can be generated based on the fusion queue information of the target fusion history queue, and the corresponding target intermediate signal can be generated based on the target original signal, and the corresponding target data sequence and its sequence number can be generated based on the target intermediate signal; the corresponding residual components can be generated based on the target original signal and the target data sequence, and if the residual components meet the decomposition conditions, the residual components are updated to the target original signal, and the execution of the step of generating the corresponding target intermediate signal based on the target original signal is returned, until the obtained residual components do not meet the decomposition conditions, resulting in multiple sequentially arranged target data sequences and target residual components, thus completing the empirical mode decomposition of the fusion queue information of the target fusion history queue; among them, the target residual components are the residual components that do not meet the decomposition conditions.

[0079] Optionally, the process of performing empirical mode decomposition on the fusion queue information of the target fusion history queue to obtain multiple target data sequences and target residual components can be specifically as follows: Generate the original target signal based on the fusion queue information of the target fusion history queue; extract all local extrema from the original target signal; generate an upper envelope based on the local maxima of each local extrema, and a lower envelope based on the local minima of each local extrema; generate a mean envelope based on the upper and lower envelopes, and generate an intermediate target signal based on the original target signal and the mean envelope; determine the next target based on the intermediate target signal and the mean envelope. The intermediate signal is used to generate a target data sequence and its sequence number based on the next target intermediate signal. The target mean envelope is determined based on the target intermediate signal. The residual component is determined based on the original target signal and the next target intermediate signal. It is determined whether the target intermediate signal meets the data sequence conditions. If the target intermediate signal meets the data sequence conditions, the residual component is updated to the original target signal, and the process of extracting all local extrema from the original target signal is returned. If the target intermediate signal does not meet the data sequence conditions, the residual component is determined as the target residual component, and the empirical mode decomposition processing of the fusion queue information of the target fusion history queue is completed.

[0080] It should be noted that if the target intermediate signal satisfies the data sequence conditions, the residual component obtained based on the target intermediate signal can be considered to satisfy the decomposition conditions, and the next round of decomposition can continue. If the target intermediate signal does not satisfy the data sequence conditions, the residual component obtained based on the target intermediate signal can be considered to not satisfy the decomposition conditions, and the next round of decomposition cannot continue. In this case, the final residual component (target residual component) can be determined based on the obtained residual component. The data sequence conditions can be: the number of intersections between local extrema and zeros is equal or the difference is no greater than 1 throughout the entire time range; at the same time, the average value of the two envelope curves formed by the local maxima and local minima should be 0 at any point.

[0081] In some embodiments, the process of determining whether the target intermediate signal satisfies the data sequence condition can be as follows: extract all local extrema and all zero intersections from the target intermediate signal; determine whether the difference in the number of all extracted local extrema is 0 or less than or equal to 1; if the difference is 0 or less than or equal to 1, further generate an upper envelope for each local maximum point among all local extrema and a lower envelope for each local minimum point among all local extrema, and determine whether the average value at any point of the mean envelope generated based on the generated upper and lower envelopes is 0. If so, it is determined that the target intermediate signal satisfies the data sequence condition, and the residual component obtained based on the target intermediate signal satisfies the decomposition condition; otherwise, it is determined that the target intermediate signal does not satisfy the data sequence condition, and the residual component obtained based on the target intermediate signal does not satisfy the decomposition condition.

[0082] In practical applications, the target historical time in the fusion queue information of the target fusion historical queue can be used as the horizontal axis, and the number of people queuing for each target historical time can be used as the vertical axis to generate the corresponding target original signal; all local extreme points are extracted from the target original signal, and the local maximum points in each local extreme point are connected to form the upper envelope, and the local minimum points in each local extreme point are connected to form the lower envelope; the mean of the upper envelope and the lower envelope are calculated to obtain the mean envelope of the current target original signal, wherein the calculation method of the mean envelope is as shown in formula (1); the difference between the target original signal and the mean envelope is calculated to obtain the target intermediate signal, wherein the calculation method of the target intermediate signal is as shown in formula (2); the upper envelope and the lower envelope of the target intermediate signal are extracted, and the mean envelope of the target intermediate signal is calculated based on the upper envelope and the lower envelope of the target intermediate signal; the difference between the target intermediate signal and its mean envelope is calculated to obtain the target mean envelope, and the target mean envelope is then connected to the target mean envelope. The network is used as the mean envelope of the next target intermediate signal to calculate the difference between the mean envelope of the target intermediate signal and the next target intermediate signal, so as to obtain the next target intermediate signal; the next target intermediate signal is used as the target data sequence, and the corresponding sequence number is marked for the target data sequence. The calculation method of the next target intermediate signal is as shown in formula (3); the difference between the target original signal and the next target intermediate signal is calculated, that is, the difference between the target original signal and the target data sequence is calculated to obtain the residual component. The calculation formula of the residual component is as shown in formula (4); if the residual component meets the decomposition condition, the residual component is updated to the target original signal, and the step of extracting all local extreme points from the target original signal is continued to continue to decompose the next target data sequence until the obtained residual component does not meet the decomposition condition. The target residual component is determined according to the residual component that does not meet the decomposition condition. That is, the calculation method of the last residual component is as shown in formula (5).

[0083] (1)

[0084] in, The mean envelope, The upper envelope, This is the lower envelope.

[0085] (2)

[0086] in, The mean envelope, For the target original signal, This is the intermediate signal for the target.

[0087] (3)

[0088] in, This serves as an intermediate signal for the next target. Let this be the mean envelope of the intermediate signal of the next target. The target data sequence.

[0089] (4)

[0090] in, The residual component corresponding to the target data sequence.

[0091] (5)

[0092] in, These are the residual components that do not meet the decomposition conditions, that is, the residual components obtained from the last decomposition. This refers to the residual components obtained from the previous decomposition before the final decomposition, which is essentially the original target signal used in the final decomposition. This is the target data sequence obtained from the last decomposition.

[0093] It should be noted that when After becoming a monotonic function, the remaining... If it becomes the final residual component, then the original target signal It is the sum of all IMF components and residual components, that is, the sum of all target data sequences and target residual components, as shown in formula (6).

[0094] (6)

[0095] It should be noted that by performing Empirical Mode Decomposition (EMD) on the fusion queue information of the target fusion history queue, the fusion queue information of the target fusion history queue can be decomposed into several Intrinsic Mode Functions (IMFs) and target residual components, that is, decomposed into several target data sequences and target participation components. Furthermore, performing EMD decomposition on the fusion queue information of the target fusion history queue can also effectively remove noise in the data, improve the stability of the data, provide better data input for subsequent prediction models, and improve the accuracy of the obtained prediction results.

[0096] Furthermore, by performing a "screening" operation on the original target signal through the EMD decomposition method, the signal can be decomposed based on its own characteristics, and different frequency components in the signal can be extracted step by step, thereby achieving the stabilization of non-stationary signals. This method does not rely on preset basis functions and has strong adaptability and flexibility.

[0097] S104: For each target data sequence, determine the prediction model that matches the target data sequence, and use the prediction model that matches the target data sequence to predict the airport security passenger flow based on the target data sequence or the target data sequence and the target residual component, and obtain the target prediction result.

[0098] In this embodiment of the application, historical airport security check queuing data within a historical time period can be obtained, and empirical mode decomposition processing can be performed on the historical airport security check queuing data to obtain multiple data sequences and training residual components. The prediction model to be trained can be trained using each data sequence and training residual component or each data sequence to obtain multiple prediction models.

[0099] It should be noted that the historical time period can be 0.5 hours, 1 hour, 2 hours, 10 hours, 100 hours, 200 hours, etc., and this embodiment of the application does not limit it.

[0100] Optionally, the process of training multiple prediction models by using various data sequences obtained from processing historical airport security check queuing data and training residual components or various data sequences to train the prediction model to be trained can be as follows: Collect historical airport security check queuing data; wherein, the historical airport security check queuing data includes airport security passenger flow sequences, wherein the airport security passenger flow sequences include queue information of multiple historical queues; divide each historical queue into a training set and a test set; wherein, the training set includes multiple training queues, and the test set includes multiple test queues; fuse the queue information of each training queue to obtain fused queue information of the fused training queue, and fuse the queue information of each test queue... The queue information of the queues is fused to obtain the fused queue information of the fused test queue; the fused queue information of the fused training queue is processed by empirical mode decomposition to obtain multiple training data sequences and training residual components arranged in sequence; each training data sequence has a corresponding index; the fused queue information of the fused test queue is processed by empirical mode decomposition to obtain multiple test data sequences and test residual components arranged in sequence; each test data sequence has a corresponding index; for each index, the prediction model to be trained is trained using the training residual components, the training data sequence with the same index, and the test data sequence to obtain the prediction model under that index.

[0101] As one implementation of this application, the process of training a prediction model to be trained using training residual components, training data sequences with the same sequence number, and test data sequences to obtain a prediction model under that sequence number can be as follows: Input the training data sequence into the prediction model to be trained, so that the prediction model to be trained uses the training data sequence to obtain an initial prediction model; validate the initial prediction model using a test data sequence with the same sequence number as the training data sequence to obtain the goodness of fit of the initial prediction model; if the goodness of fit of the initial prediction model is greater than a preset goodness of fit, determine that the initial prediction model is the prediction model under the sequence number of the training data sequence; wherein, the training data used during prediction model training is the training data sequence; if the goodness of fit of the initial prediction model is not greater than the preset goodness of fit, input the training data sequence and training residual components into the initial prediction model, so that the initial prediction model uses the training data sequence and training residual components to obtain a prediction model under the sequence number of the training data sequence; wherein, the training data used during prediction model training is the training data sequence and training residual components.

[0102] It should be noted that the prediction model to be trained can be LSTM or XGBoost. The corresponding neural network or algorithm can be selected as the prediction model to be trained according to the actual application. This application does not limit this.

[0103] In practical applications, assuming the prediction model to be trained is LSTM, the training process for each prediction model can be found in [reference needed]. Figure 2The historical queues in the airport passenger flow sequence (i.e., the airport security check passenger flow sequence) are divided into training and testing sets. Empirical Mode Decomposition (EMD) is performed on the fused queue information of the fused training queue obtained by fusing the queue information of each training queue in the training set, resulting in n training data sequences and training residual components, i.e., n IMF components (n training data sequences) and training residual components. Similarly, EMD is performed on the fused test queue information of the fused test queue obtained by fusing the queue information of each test queue in the testing set, resulting in n test data sequences and test residual components, i.e., n IMF components (n test data sequences) and test residual components. These residual components can then be used to... The LSTM is trained using training data sequences and test sequences with the same sequence number 1 to obtain the prediction model corresponding to sequence number 1. ... The LSTM is trained using training residual components, training data sequences with the same sequence number n, and test sequences to obtain the prediction model corresponding to sequence number n. Assuming that the historical airport security queue data before the target time is processed to obtain n target data sequences, the prediction model matching each target data sequence can be determined from the n prediction models. Airport security passenger flow is predicted using each target data sequence and its matching prediction model to obtain n target prediction results. Finally, the n target prediction results are superimposed and reconstructed to obtain the final prediction result, that is, the airport security passenger flow corresponding to the target time.

[0104] In some embodiments, the number of people queuing in the same historical time from each training queue is merged to obtain the merged queue information of the initial merged training queue; the merged queue information of the initial merged training queue is cleaned and normalized to obtain the merged queue information of the merged training queue; the number of people queuing in the same historical time from each test queue is merged to obtain the merged queue information of the initial merged test queue; the merged queue information of the initial merged test queue is cleaned and normalized to obtain the merged queue information of the merged test queue.

[0105] It should be noted that data cleaning and normalization of the fusion queue information of the initial fusion training queue and the fusion queue information of the initial fusion test queue can not only remove control and error values ​​from the data, but also scale the number of people in the queue corresponding to each historical time to the range of [0, 1]. This can improve the efficiency and stability of model training when using this data for subsequent model training.

[0106] In this embodiment, the process of performing empirical mode decomposition on the fusion queue information of the fusion training queue to obtain multiple sequentially arranged training data sequences can be as follows: Using each historical time in the fusion queue information of the fusion training queue as the abscissa and the number of people queuing at each historical time as the ordinate, a corresponding original signal is generated; all local extrema are extracted from the original signal, and the local maxima points among the extracted local extrema are connected to form an upper envelope, and the local minima points among the extracted local extrema are connected to form a lower envelope; the mean of the upper and lower envelopes is calculated to obtain the mean envelope corresponding to the current original signal; the difference between the original signal and the mean envelope is calculated to obtain the intermediate signal; the upper and lower envelopes of the intermediate signal are extracted, and based on the upper and lower envelopes of the intermediate signal... Calculate the mean envelope of the intermediate signal, and calculate the difference between the intermediate signal and its mean envelope to obtain the mean envelope of the next intermediate signal. Then, calculate the difference between the mean envelope of the intermediate signal and the next intermediate signal to obtain the next intermediate signal. Use this next intermediate signal as the training data sequence and label it with a corresponding index. Calculate the difference between the original signal and the next intermediate signal, i.e., calculate the difference between the original signal and the training data sequence, to obtain the corresponding residual component. If the residual component satisfies the decomposition condition, update the original signal with the residual component and continue the step of extracting all local extrema from the original signal to continue decomposing the next training data sequence until the obtained residual component no longer satisfies the decomposition condition. Finally, determine the training residual component based on the residual component that does not satisfy the decomposition condition.

[0107] It's also worth noting that LSTM is a variant of a recurrent neural network (RNN) specifically designed for processing time-series data. Compared to traditional RNNs, LSTM effectively addresses long-term dependencies and avoids the vanishing gradient problem. The internal process of LSTM can be divided into three steps:

[0108] (1) Using input Hidden state information from the previous time step Generate three gating information and one candidate memory, namely: the forget gate. Input gate Output gate Candidate memories ;

[0109] (2) The forgetting gate controls the degree to which historical information is forgotten. Input gate controls access to historical memory Add candidate memories The degree to which forgotten memories, combined with selected memories, are updated into a new memory state. ;

[0110] (3) The output gate controls the new memory state to become a hidden state. The degree of [something]. New memories and hidden states are the output at the current moment.

[0111] In the above steps This represents the input at time t. The calculation process for each gate information is as follows:

[0112] Forgotten Gate :

[0113] (7)

[0114] Input gate :

[0115] (8)

[0116] Output gate :

[0117] (9)

[0118] Candidate Memory :

[0119] (10)

[0120] Current cell state :

[0121] (11)

[0122] Output hidden layer :

[0123] (12)

[0124] In formulas (1)-(6), , , , , These represent the forget gate, input gate, output gate, candidate memory, and current unit memory, respectively. These are the weight matrix representations in LSTM. This is the bias vector representation in LSTM.

[0125] As one implementation of this application, assuming the prediction model to be trained is LSTM, the training data sequence can be used as the input to the LSTM. This allows the LSTM to make predictions using the input training data sequence according to the internal process shown above, obtain the prediction results of the training data sequence, and adjust the parameters of the LSTM with the training objective of the prediction results of the training data sequence approaching the actual results of the training data sequence, until the LSTM reaches convergence, thus obtaining the initial prediction model under the sequence number of the training data sequence.

[0126] It should also be noted that XGBoost is an ensemble learning algorithm that improves prediction performance by constructing and combining multiple weak classifiers (usually decision trees). It uses additive training, learning one CART tree in each iteration to fit the residuals between the predictions of the previous k-1 trees and the true values ​​of the training samples. The final result is obtained by summing the results of all k regression trees. Assuming a total of K regression trees are constructed, then for the i-th sample... The predicted value is shown in formula (13):

[0127] (13)

[0128] Where K represents the total number of trees, Represents the k-th tree, Representative sample x i The prediction results.

[0129] As another implementation of this application, assuming the prediction model to be trained is XGBoost, the training data sequence can be used as the input x of XGBoost. i In order to enable XGBoost to make predictions using the input training data sequence according to the training process and formula (13) shown above, obtain the prediction results of the training data sequence, and take the prediction results of the training data sequence as close as possible to the actual results of the training data sequence as the training objective, adjust the parameters of XGBoost until XGBoost reaches convergence, and obtain the initial prediction model under the sequence number of the training data sequence.

[0130] It should be noted that the method of performing empirical mode decomposition on the fusion queue information of the fusion test queue is the same as the method of performing empirical mode decomposition on the fusion queue information of the fusion training queue. Please refer to the corresponding content above. This embodiment of the application will not repeat the details here.

[0131] In the specific execution step S104, after obtaining multiple target data sequences, for each target data sequence, the prediction model matching the target data sequence can be selected from various prediction models according to the sequence number corresponding to the target data sequence, so that the prediction model matching the target data sequence can be used to predict the airport security passenger flow based on the target data sequence or the target data sequence and the target residual component, and the target prediction result can be obtained.

[0132] Optionally, for each target data sequence, the process of determining a prediction model that matches the target data sequence and using the prediction model that matches the target data sequence to predict airport security passenger flow based on the target data sequence or the target data sequence and target residual components to obtain the target prediction result can be as follows: For each target data sequence, the prediction model with the same sequence number as the target data sequence is determined as the prediction model that matches the target data sequence; it is determined whether the training data used by the prediction model during training is a training data sequence, or whether it is a training data sequence and training residual components; if the training data used by the prediction model during training is a training data sequence, the target data sequence is input into the prediction model that matches it, so that the prediction model that matches the target data sequence makes a prediction based on the target data sequence to obtain the target prediction result corresponding to the target data sequence; if the training data used by the prediction model during training is a training data sequence and training residual components, the target data sequence and target residual components are input into the prediction model that matches it, so that the prediction model that matches the target data sequence makes a prediction based on the target data sequence and target residual components to obtain the target prediction result corresponding to the target data sequence.

[0133] S105: Determine the airport security passenger flow corresponding to the target time based on the prediction results of each target.

[0134] In the specific execution of step S105, after obtaining multiple target prediction results, the prediction results can be summed to obtain the airport security passenger flow corresponding to the target time, or the prediction results can be weighted and averaged to obtain the airport security passenger flow corresponding to the target time.

[0135] Furthermore, in this embodiment of the application, after obtaining the airport security passenger flow corresponding to the target time, the airport security passenger flow corresponding to the target time can be displayed to airport management personnel through a user-interactive visual interface, so that airport management personnel can monitor the airport security passenger flow at various times in real time and make resource allocation and decision support based on the detected airport security passenger flow.

[0136] This application provides a method for predicting airport security passenger flow. The method involves determining the target time for passenger flow prediction and acquiring historical airport security queue data prior to the target time. This historical airport security queue data includes queue information from multiple historical queues. The queue information from each historical queue is then fused to obtain fused queue information for the target fused historical queue. Empirical mode decomposition (EMD) is applied to the fused queue information to obtain multiple target data sequences and target residual components. For each target data sequence, a prediction model matching the target data sequence is determined. The prediction model is then used to predict airport security passenger flow based on the target data sequence or the target data sequence and the target residual components, yielding the target prediction result. Each prediction model is trained using the data sequences obtained from processing historical airport security queue data and the training residual components or the data sequences to be trained. The airport security passenger flow corresponding to the target time is determined based on the target prediction results. Therefore, the technical solution provided in this application can effectively decompose the original non-stationary time series data by performing empirical mode decomposition on the target historical airport security check data, remove noise from the data, and extract multiple target data sequences and target residual components. Then, each target data sequence or each target data sequence and target residual component is input into a matching prediction model for prediction. By combining the time series prediction capability of the prediction model, the prediction accuracy of complex non-stationary time series data can be effectively improved, that is, the accuracy of predicting airport security check passenger flow can be improved.

[0137] Furthermore, in the embodiments of this application, the applicant has found that, through the training method of the prediction model provided in this application, prediction models corresponding to two time granularities with historical time periods of 0.5 hours and 1 hour can be constructed respectively. Then, the prediction models trained by the EMD+LSTM or EMD+XGBoost training methods provided in this application are compared and analyzed with the prediction models trained by the single LSTM training method and the prediction models trained by the XGBoost training method. The specific comparison and analysis results are shown in Table 1.

[0138] Table 1:

[0139]

[0140] It should be noted that, as shown in Table 1, the EMD-LSTM model outperforms the prediction models obtained using single LSTM training and XGBoost training methods at time granularities of 0.5 hours and 1 hour. This demonstrates that the Empirical Mode Decomposition algorithm can effectively reduce noise, optimize the prediction performance of the prediction model, and improve the accuracy of predicting airport security passenger flow, thereby providing scientific decision support for airport managers.

[0141] Therefore, this application, after summarizing historical airport security queue data, found that it exhibits obvious time-series characteristics. To address the noise problem in non-stationary historical airport security queue data, this application introduces the Empirical Mode Decomposition (EMD) algorithm and combines it with a corresponding prediction model to be trained (such as LSTM) to construct a corresponding prediction model, which can effectively optimize the performance of the prediction model when processing noisy data.

[0142] Furthermore, as can be seen from Table 1, at different time granularities, the EMD-LSTM model outperforms the prediction models trained using a single LSTM training method or XGBoost training method in terms of prediction accuracy at 0.5 hours and 1 hour.

[0143] In summary, the airport security passenger flow prediction method provided in this application utilizes the EMD algorithm to decompose unstable historical airport security queue data, reducing noise in the data and providing higher-quality data input for subsequent prediction models. This improves the prediction accuracy of the prediction models, thereby enhancing the accuracy of airport security passenger flow prediction, providing data support for airport security planning and resource scheduling, and optimizing the passenger security experience.

[0144] Based on the airport security passenger flow prediction method provided in the above embodiments of this application, correspondingly, the present application also provides an airport security passenger flow prediction system, such as... Figure 3 As shown, the airport security passenger flow prediction system includes:

[0145] The data acquisition module 31 is used to determine the target time for the airport passenger flow forecast and to acquire the target historical airport security check queue data before the target time; wherein, the target historical airport security check queue data includes queue information of multiple target historical queues;

[0146] Data preprocessing module 32 is used to fuse the queue information of each target historical queue to obtain the fused queue information of the target fused historical queue;

[0147] EMD decomposition module 33 is used to perform empirical mode decomposition on the fusion queue information of the target fusion history queue to obtain multiple target data sequences and target residual components.

[0148] Prediction module 34 is used to determine the prediction model matching the target data sequence for each target data sequence, and to predict the airport security passenger flow based on the target data sequence or the target data sequence and the target residual component using the prediction model matching the target data sequence, so as to obtain the target prediction result; wherein, each prediction model is obtained by the training module using the data sequences obtained by processing historical airport security queue data and the training residual component or the prediction model to be trained by the training module.

[0149] The result reconstruction module 35 is used to determine the airport security passenger flow corresponding to the target time based on the prediction results of each target.

[0150] The specific principles and execution processes of each unit in the airport security passenger flow prediction system disclosed in the above embodiments of this application are the same as those in the airport security passenger flow prediction method disclosed in the above embodiments of this application. Please refer to the corresponding parts in the airport security passenger flow prediction method disclosed in the above embodiments of this application, and they will not be repeated here.

[0151] This application provides an airport security check passenger flow prediction system. The system determines the target time for airport passenger flow prediction and acquires historical airport security check queuing data prior to the target time. The historical airport security check queuing data includes queue information from multiple historical queues. The system merges the queue information from each historical queue to obtain fused queue information for the fused historical queue. Empirical mode decomposition (EMD) is applied to the fused queue information to obtain multiple target data sequences and target residual components. For each target data sequence, a prediction model matching the target data sequence is determined. The system then uses this prediction model to predict airport security check passenger flow based on the target data sequence or the target data sequence and the target residual components, obtaining the target prediction result. Each prediction model is trained using the data sequences obtained from processing historical airport security check queuing data and the training residual components or the data sequences to be trained. The system determines the airport security check passenger flow corresponding to the target time based on each target prediction result. Therefore, the technical solution provided in this application can effectively decompose the original non-stationary time series data by performing empirical mode decomposition on the target historical airport security check data, remove noise from the data, and extract multiple target data sequences and target residual components. Then, each target data sequence or each target data sequence or target residual component is input into a matching prediction model for prediction. By combining the time series prediction capability of the prediction model, the prediction accuracy of complex non-stationary time series data can be effectively improved, that is, the accuracy of predicting airport security check passenger flow can be improved.

[0152] Optionally, the queue information of the target historical queue includes multiple consecutive target historical times and the number of people queuing for each target historical time. The data preprocessing module includes:

[0153] The fusion unit is used to merge the number of people queuing in the same target historical time in each target historical queue to obtain the fusion queue information of the initial fusion historical queue.

[0154] The first preprocessing unit is used to preprocess the fusion queue information of the initial fusion history queue to obtain the fusion queue information of the target fusion history queue.

[0155] Optional, the EMD decomposition module includes:

[0156] The target raw signal generation unit is used to generate the target raw signal based on the fusion queue information of the target fusion history queue;

[0157] The first extraction unit is used to extract all local extrema from the original target signal;

[0158] The first envelope generation unit is used to generate an upper envelope based on local maxima among the local extrema, and to generate a lower envelope based on local minima among the local extrema.

[0159] The target intermediate signal generation unit is used to generate a mean envelope based on the upper envelope and the lower envelope, and to generate a target intermediate signal based on the original target signal and the mean envelope.

[0160] The target data sequence generation unit is used to determine the next target intermediate signal based on the target intermediate signal and the target mean envelope, and to generate the target data sequence and its sequence number based on the next target intermediate signal; wherein, the target mean envelope is determined based on the target intermediate signal;

[0161] The first determining unit is used to determine the residual component based on the original target signal and the next target intermediate signal;

[0162] The judgment unit is used to determine whether the target intermediate signal meets the data sequence conditions.

[0163] The return execution unit is used to update the residual components to the original target signal if the target intermediate signal meets the data sequence conditions, and then return to the execution first extraction unit;

[0164] The second determining unit is used to determine the residual component as the target residual component if the target intermediate signal does not meet the data sequence conditions, and to determine the completion of the empirical mode decomposition processing of the fusion queue information of the target fusion history queue.

[0165] Optional prediction modules include:

[0166] The matching unit is used to identify the prediction model with the same index as the target data sequence among the various prediction models for each target data sequence as the prediction model that matches the target data sequence.

[0167] The judgment unit is used to determine whether the training data used by the prediction model during the training process is a training data sequence, or whether it is a training data sequence and training residual components.

[0168] The first prediction unit is used to input the target data sequence into the prediction model that matches it if the training data used by the prediction model during the training process is a training data sequence, so that the prediction model that matches the target data sequence can make a prediction based on the target data sequence and obtain the target prediction result corresponding to the target data sequence.

[0169] The second prediction unit is used to input the target data sequence and target residual components into the prediction model that matches the target data sequence if the training data used by the prediction model during the training process is a training data sequence and a training residual component. The prediction model that matches the target data sequence makes a prediction based on the target data sequence and the target residual component, and obtains the target prediction result corresponding to the target data sequence.

[0170] Optional, training modules include:

[0171] The data collection unit is used to collect historical airport security check queue data; the historical airport security check queue data includes airport security passenger flow sequences, and the airport security passenger flow sequences include queue information of multiple historical queues;

[0172] A partitioning unit is used to divide each historical queue into a training set and a test set; the training set includes multiple training queues, and the test set includes multiple test queues.

[0173] The second preprocessing unit is used to fuse the queue information of each training queue to obtain the fused queue information of the fused training queue, and to fuse the queue information of each test queue to obtain the fused queue information of the fused test queue.

[0174] The first EMD decomposition unit is used to perform empirical mode decomposition on the fusion queue information of the fusion training queue to obtain multiple training data sequences and training residual components arranged in sequence; wherein, each training data sequence has a corresponding sequence number.

[0175] The second EMD decomposition unit is used to perform empirical mode decomposition on the fusion queue information of the fusion test queue to obtain multiple test data sequences and test residual components arranged in sequence; wherein, each test data sequence has a corresponding sequence number;

[0176] The training unit is used to train the prediction model to be trained for each index using the training residual components, the training data sequence with the same index, and the test data sequence, so as to obtain the prediction model under that index.

[0177] Optional training units include:

[0178] The first training subunit is used to input the training data sequence into the prediction model to be trained, so that the prediction model to be trained can be trained using the training data sequence to obtain the initial prediction model.

[0179] The validation subunit is used to validate the initial prediction model using a test data sequence with the same sequence number as the training data sequence, and to obtain the fit of the initial prediction model.

[0180] The third determining unit is used to determine the initial prediction model as the prediction model under the sequence number of the training data sequence if the goodness of fit of the initial prediction model is greater than the preset goodness of fit; wherein, the training data used during the training of the prediction model is the training data sequence.

[0181] The second training subunit is used to input the training data sequence and training residual components into the initial prediction model if the fit of the initial prediction model is not greater than the preset fit, so that the initial prediction model can be trained using the training data sequence and training residual components to obtain the prediction model under the sequence number of the training data sequence; wherein, the training data used during the training of the prediction model are the training data sequence and training residual components.

[0182] This application also provides a storage medium storing program instructions, which, when loaded and executed by a processor, implement any of the above-described embodiments of the airport security passenger flow prediction method.

[0183] This application also provides an electronic device, such as Figure 4 As shown, the device includes a processor 401 and a memory 402, which are connected via a bus; the memory stores program instructions; the processor calls the program instructions in the memory to execute any of the above-described embodiments of the airport security passenger flow prediction method.

[0184] The processor mentioned in this article can be the terminal's CPU, an integrated MCU within the terminal, or a combination of a CPU and an MCU. Furthermore, the processor contains a kernel that retrieves the corresponding program from memory; one or more kernels can be configured.

[0185] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0186] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for system or system embodiments, since they are basically similar to method embodiments, the description is relatively simple, and relevant parts can be referred to the descriptions in the method embodiments. The systems and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0187] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0188] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0189] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for predicting passenger flow at airport security checkpoints, characterized in that, The method includes: Determine the target time for airport passenger traffic forecasting and obtain the target historical airport security queue data prior to the target time; wherein, the target historical airport security queue data includes queue information of multiple target historical queues; The queue information of each target historical queue is merged to obtain the merged queue information of the target merged historical queue; Empirical mode decomposition is performed on the fusion queue information of the target fusion history queue to obtain multiple target data sequences and target residual components; For each target data sequence, a prediction model matching the target data sequence is determined, and airport security passenger flow is predicted based on the target data sequence or the target data sequence and the target residual component using the prediction model matching the target data sequence to obtain the target prediction result; wherein, each prediction model is obtained by training the prediction model to be trained using each data sequence obtained by processing historical airport security queue data and training residual component or each data sequence. The airport security passenger flow corresponding to the target time is determined based on the prediction results of each target.

2. The method according to claim 1, characterized in that, The queue information of the target history queue includes multiple consecutive target history times and the number of people in the queue corresponding to each target history time; The step of fusing the queue information of each of the target historical queues to obtain the fused queue information of the target fused historical queue includes: The number of people queuing in each of the target historical queues belonging to the same target historical time is merged to obtain the merged queue information of the initial merged historical queue; The fusion queue information of the initial fusion history queue is preprocessed to obtain the fusion queue information of the target fusion history queue.

3. The method according to claim 1, characterized in that, The empirical mode decomposition process is performed on the fusion queue information of the target fusion history queue to obtain multiple target data sequences and target residual components, including: The original target signal is generated based on the fusion queue information of the target fusion history queue; Extract all local extrema from the original target signal; An upper envelope is generated based on the local maxima among the local extrema, and a lower envelope is generated based on the local minima among the local extrema. A mean envelope is generated based on the upper envelope and the lower envelope, and a target intermediate signal is generated based on the target original signal and the mean envelope. Based on the target intermediate signal and the target mean envelope, the next target intermediate signal is determined, and a target data sequence and its sequence number are generated based on the next target intermediate signal; wherein, the target mean envelope is determined based on the target intermediate signal; The residual component is determined based on the original target signal and the intermediate signal of the next target; Determine whether the target intermediate signal satisfies the data sequence conditions; If the target intermediate signal satisfies the data sequence condition, the residual component is updated to the target original signal, and the process returns to the step of extracting all local extrema from the target original signal. If the target intermediate signal does not meet the data sequence conditions, the target residual component is determined based on the residual component, and the empirical mode decomposition process of the fusion queue information of the target fusion history queue is completed.

4. The method according to claim 3, characterized in that, For each target data sequence, a prediction model matching the target data sequence is determined, and airport security passenger flow is predicted based on the target data sequence or the target data sequence and the target residual component using the prediction model matching the target data sequence, to obtain the target prediction result, including: For each target data sequence, the prediction model with the same index as the target data sequence is determined as the prediction model that matches the target data sequence. Determine whether the training data used by the prediction model during the training process is a training data sequence, or whether it is a training data sequence and training residual components; If the training data used by the prediction model during the training process is a training data sequence, the target data sequence is input into the prediction model that matches it, so that the prediction model that matches the target data sequence makes a prediction based on the target data sequence and obtains the target prediction result corresponding to the target data sequence. If the training data used by the prediction model during training is a training data sequence and a training residual component, the target data sequence and the target residual component are input into the prediction model that matches it, so that the prediction model that matches the target data sequence makes a prediction based on the target data sequence and the target residual component, and obtains the target prediction result corresponding to the target data sequence.

5. The method according to claim 1, characterized in that, The process involves processing historical airport security queuing data to obtain various data sequences and training residual components, or using various data sequences to train a prediction model to be trained, resulting in multiple prediction models, including: Collect historical airport security check queue data; wherein, the historical airport security check queue data includes airport security check passenger flow sequences, and the airport security check passenger flow sequences include queue information of multiple historical queues; Each of the historical queues is divided into a training set and a test set; wherein the training set includes multiple training queues and the test set includes multiple test queues. The queue information of each training queue is merged to obtain the merged queue information of the training queue, and the queue information of each test queue is merged to obtain the merged queue information of the test queue. The fusion queue information of the fusion training queue is processed by empirical mode decomposition to obtain multiple training data sequences and training residual components arranged in sequence; wherein, each training data sequence has a corresponding sequence number; The fusion queue information of the fusion test queue is processed by empirical mode decomposition to obtain multiple test data sequences and test residual components arranged in sequence; wherein, each test data sequence has a corresponding sequence number; For each index, the prediction model to be trained is trained using the training residual components, the training data sequence with the same index, and the test data sequence to obtain the prediction model under that index.

6. The method according to claim 5, characterized in that, The step of training the prediction model to be trained using the training residual components, the training data sequence with the same sequence number, and the test data sequence to obtain the prediction model under that sequence number includes: The training data sequence is input into the prediction model to be trained, so that the prediction model to be trained can be trained using the training data sequence to obtain an initial prediction model. The initial prediction model is validated using a test data sequence with the same sequence number as the training data sequence to obtain the goodness of fit of the initial prediction model; If the goodness of fit of the initial prediction model is greater than the preset goodness of fit, the initial prediction model is determined to be the prediction model under the sequence number of the training data sequence; wherein, the training data used during the training of the prediction model is the training data sequence. If the goodness of fit of the initial prediction model is not greater than the preset goodness of fit, the training data sequence and the training residual components are input into the initial prediction model, so that the initial prediction model is trained using the training data sequence and the training residual components to obtain the prediction model under the sequence number of the training data sequence; wherein, the training data used when the prediction model is trained are the training data sequence and the training residual components.

7. An airport security check passenger flow prediction system, characterized in that, The system includes: The data acquisition module is used to determine the target time for airport passenger flow prediction and to acquire the target historical airport security check queue data before the target time; wherein, the target historical airport security check queue data includes queue information of multiple target historical queues; The data preprocessing module is used to fuse the queue information of each target historical queue to obtain the fused queue information of the target fused historical queue; The EMD decomposition module is used to perform empirical mode decomposition processing on the fusion queue information of the target fusion history queue to obtain multiple target data sequences and target residual components. The prediction module is used to determine a prediction model matching the target data sequence for each target data sequence, and to predict airport security passenger flow based on the target data sequence or the target data sequence and the target residual component using the prediction model matching the target data sequence, thereby obtaining a target prediction result; wherein, each prediction model is obtained by the training module using each data sequence obtained by processing historical airport security queue data and training residual components or each data sequence to be trained prediction model; The result reconstruction module is used to determine the airport security passenger flow corresponding to the target time based on the prediction results of each target.

8. The system according to claim 7, characterized in that, The queue information of the target historical queue includes multiple consecutive target historical times and the number of people queuing for each target historical time. The data preprocessing module includes: The fusion unit is used to merge the number of people queuing in the same target historical time in each of the target historical queues to obtain the fusion queue information of the initial fusion historical queue. The preprocessing unit is used to preprocess the fusion queue information of the initial fusion history queue to obtain the fusion queue information of the target fusion history queue.

9. An electronic device, characterized in that, include: A processor and a memory are connected via a bus; wherein the processor is used to call and execute a program stored in the memory; The memory is used to store a program for implementing an airport security passenger flow prediction method as described in any one of claims 1-6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions for performing an airport security passenger flow prediction method as described in any one of claims 1-6.

Images