Model training method, passenger flow prediction method, equipment and medium

Abstract

The invention discloses a model training method, a passenger flow prediction method, equipment and a medium, and relates to the technical field of financial science and technology. The model training method comprises the following steps: acquiring branch queuing passenger flow volume sample data; and inputting the branch queuing passenger flow volume sample data into a preset deep learning model, and performing sequence time change feature analysis and data time dependency analysis to obtain a branch queuing passenger flow prediction model, thereby predicting the branch queuing passenger flow volume through the branch queuing passenger flow prediction model. According to the technical scheme of the embodiment of the invention, the prediction precision and efficiency of the queuing passenger flow can be improved.

Description

Technical Field

[0001] The present invention relates to the field of financial technology, and in particular to a model training method, a passenger flow prediction method, a device, and a medium. Background Art

[0002] As a crucial gateway to financial services, the volume and volatility of customer traffic at bank branches directly impacts service quality and customer satisfaction. Therefore, accurately predicting bank branch queue flow is crucial for optimizing service processes and improving operational efficiency.

[0003] With the rapid development of big data technology, the scale and complexity of customer flow data continue to increase. This makes bank branch customer flow data characterized by time dependence, multiple periods, randomness, and nonlinearity. Traditional customer flow forecasting methods, often based on statistical or machine learning models, struggle to achieve ideal accuracy and efficiency when processing large-scale, high-dimensional customer flow data. Summary of the Invention

[0004] The present invention provides a model training method, a passenger flow prediction method, a device and a medium to solve the problems of poor prediction accuracy and efficiency of existing queuing passenger flow prediction methods.

[0005] According to one aspect of the present invention, a model training method is provided, comprising:

[0006] Obtain sample data on customer flow in queues at outlets;

[0007] Input the sample data of customer flow in queues at outlets into the preset deep learning model, perform analysis on the time variation characteristics of the sequence and the time dependency between the data, and obtain the customer flow prediction model for the outlet queues;

[0008] Among them, the outlet queue passenger flow prediction model is used to predict the outlet queue passenger flow.

[0009] According to another aspect of the present invention, a method for predicting customer flow in a network station queue is provided, comprising:

[0010] Obtaining data related to customer flow prediction at outlets;

[0011] Through the outlet queue passenger flow prediction model, the outlet queue passenger flow prediction related data is analyzed to obtain the outlet queue estimated passenger flow;

[0012] The network queuing passenger flow prediction model is a model trained by the model training method in any embodiment of the present invention.

[0013] According to another aspect of the present invention, there is provided a model training device, comprising:

[0014] Sample data acquisition module, used to obtain sample data of customer flow in queues at outlets;

[0015] The model training model is used to input the sample data of the customer flow in the queue of the outlet into the preset deep learning model, perform sequence time variation feature analysis and time dependency analysis between the data, and obtain the customer flow prediction model of the queue of the outlet;

[0016] Among them, the outlet queue passenger flow prediction model is used to predict the outlet queue passenger flow.

[0017] According to another aspect of the present invention, there is provided a device for predicting customer flow in a network queue, comprising:

[0018] Real-time data acquisition module, used to obtain data related to the prediction of customer flow in queues at outlets;

[0019] The passenger flow prediction module is used to analyze the network queue passenger flow prediction related data through the network queue passenger flow prediction model to obtain the estimated network queue passenger flow; wherein the network queue passenger flow prediction model is a model trained by the model training method in any embodiment.

[0020] According to another aspect of the present invention, an electronic device is provided, comprising:

[0021] at least one processor; and

[0022] a memory communicatively connected to the at least one processor; wherein,

[0023] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the model training method described in any embodiment of the present invention, or execute the network queuing passenger flow prediction method described in an embodiment of the present invention.

[0024] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the model training method described in any embodiment of the present invention when executed, or to execute the network queue passenger flow prediction method described in an embodiment of the present invention.

[0025] The technical solution of an embodiment of the present invention obtains sample data of customer flow in queues at outlets, inputs the sample data into a preset deep learning model, performs analysis of the time-varying characteristics of the sequence and the temporal dependency between data, and obtains an outlet queue customer flow prediction model. Because the modeling of actual queue customer flow involves multiple variables, the training of the preset deep learning model's analysis of the time-varying characteristics of the sequence based on the sample data of customer flow in queues at outlets can consider the interactions between multiple variables and their associated characteristics with events to learn the periodicity between sequences and the temporal variation between cycles. The training of the analysis of the temporal dependency between data can enhance the ability to capture dependencies between data. Compared with traditional neural network models, it can also effectively reduce computational complexity, solving the problems of poor prediction accuracy and efficiency of existing customer flow prediction methods, and can improve the prediction accuracy and efficiency of customer flow.

[0026] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present invention, nor is it intended to limit the scope of the present invention. Other features of the present invention will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0028] Figure 1 A flowchart of a model training method provided in Example 1 of the present invention;

[0029] Figure 2 A flowchart of a model training method provided in Example 2 of the present invention;

[0030] Figure 3 A logical diagram of queuing passenger flow prediction provided by the third embodiment of the present invention;

[0031] Figure 4 This is an architectural diagram of a two-dimensional feature module provided in Example 3 of the present invention;

[0032] Figure 5 A structural diagram of a two-dimensional feature module provided in Example 3 of the present invention;

[0033] Figure 6 A schematic structural diagram of a model training device provided in a fourth embodiment of the present invention;

[0034] Figure 7A schematic structural diagram of an electronic device that can be used to implement an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0035] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0036] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0037] Example 1

[0038] Figure 1 This is a flow chart of a model training method provided in the first embodiment of the present invention. This embodiment is applicable to the situation of accurately predicting the passenger flow in queues. The method can be executed by a model training device. The model training device can be implemented in the form of hardware and / or software. The model training device can be configured in an electronic device. The electronic device can include but is not limited to a server or a computer. Figure 1 As shown, the method includes:

[0039] Step 110: Obtain sample data of customer flow in queues at outlets.

[0040] The sample branch queue flow data may be the result of data cleaning of historical branch queue flow data. For example, the historical branch queue flow data may include, but is not limited to, historical queue flow data for bank branches in various regions, as well as factors influencing queue flow. Factors influencing queue flow may include holidays, promotional events, weather conditions, and traffic conditions. Optionally, influencing factors may be obtained through market research using expert analysis methods and / or by searching third-party data sources.

[0041] In the embodiment of the present invention, historical data of customer flow in queues at outlets may be obtained, and then data preprocessing may be performed on the historical data of customer flow in queues to determine sample data of customer flow in queues at outlets.

[0042] Step 120: Input the sample data of the customer flow in queue at the outlet into a preset deep learning model, perform sequence time variation feature analysis and time dependency analysis between data, and obtain an outlet queue customer flow prediction model.

[0043] The preset deep learning model can be a pre-set neural network model. Sequential time variation feature analysis can be an operation that converts a one-dimensional time series into a two-dimensional space and then extracts and analyzes features. Inter-data temporal dependency analysis can be an analysis of the temporal dependencies between sequence data. For example, the temporal dependencies between data can be captured using a self-attention transformer (autoformer model) with an autoregressive mechanism within the preset deep learning model. The outlet queue flow prediction model can be used to predict outlet queue flow.

[0044] In an embodiment of the present invention, the sample data of the queuing passenger flow at the outlets can be input into a preset deep learning model to train the preset deep learning model for serial time change feature analysis and time dependency analysis between data based on the sample data of the queuing passenger flow at the outlets, and the trained model can be used as the queuing passenger flow prediction model for predicting the queuing passenger flow at the outlets through the queuing passenger flow prediction model.

[0045] The technical solution of an embodiment of the present invention obtains sample data of customer flow in queues at outlets, inputs the sample data into a preset deep learning model, performs analysis of the time-varying characteristics of the sequence and the temporal dependency between data, and obtains an outlet queue flow prediction model. Because the modeling of actual queue flow involves multiple variables, the training of the preset deep learning model for the analysis of the time-varying characteristics of the sequence based on the sample data of customer flow in queues at outlets can consider the interactions between multiple variables and their associated characteristics with events to learn the periodicity between sequences and the temporal variation between cycles. The training of the analysis of the temporal dependency between data can improve the ability to capture the dependencies between data. Compared with traditional neural network models, it can also effectively reduce the computational complexity, solve the problems of poor prediction accuracy and efficiency of existing queue flow prediction methods, and can improve the prediction accuracy and efficiency of queue flow.

[0046] Example 2

[0047] Figure 2 This is a flow chart of a model training method provided by the second embodiment of the present invention. This embodiment is specific based on the above embodiment and provides a specific optional implementation method for obtaining sample data of customer flow in queues at outlets. Figure 2As shown, the method includes:

[0048] Step 210: Obtain historical data on customer queuing flow at at least one business outlet.

[0049] The historical data of customer flow in queues at outlets may be data related to customer flow in queues at outlets.

[0050] In an embodiment of the present invention, historical data on customer queuing volume of at least one bank branch may be collected.

[0051] Step 220: Clean and pre-process the historical data of customer flow in queues at the outlets to obtain sample data of customer flow in queues at the outlets.

[0052] In an embodiment of the present invention, the historical data of customer flow in queues at outlets may be cleaned and preprocessed, and the cleaned data may be used as sample data of customer flow in queues at outlets.

[0053] Step 230: Input the sample data of the customer flow in queue at the outlet into a preset deep learning model, perform sequence time variation feature analysis and time dependency analysis between data, and obtain an outlet queue customer flow prediction model.

[0054] Optionally, errors, duplications or outliers in the historical data of outlet queue passenger flow can be eliminated, and the preprocessed historical data of outlet queue passenger flow can be divided into a training set, a validation set and a test set, and then the training set is used as the outlet queue passenger flow sample data. After the preset deep learning model is trained with the outlet queue passenger flow sample data to perform sequence time change feature analysis and time dependency analysis between data, the model is optimized with the validation set and the test set to obtain the final outlet queue passenger flow prediction model.

[0055] In an optional embodiment of the present invention, the sample data of the queuing passenger flow at the outlet is input into a preset deep learning model, and a serial time change feature analysis and a time dependency analysis between data are performed to obtain an outlet queuing passenger flow prediction model. This may include: inputting the sample data of the queuing passenger flow at the outlet into a preset deep learning model, and performing at least one serial time change feature analysis and a time dependency analysis between data in a decoding module and an encoding module of the preset deep learning model to obtain an outlet queuing passenger flow prediction model.

[0056] In an embodiment of the present invention, the sample data of the queuing passenger flow at the outlets can be input into a preset deep learning model, so that based on the sample data of the queuing passenger flow at the outlets, the decoding module and the encoding module of the preset deep learning model are respectively subjected to at least one sequence time change feature analysis, and the time dependency analysis between the data is performed to obtain the outlet queuing passenger flow prediction model.

[0057] In an optional embodiment of the present invention, performing at least one sequence time change feature analysis in the decoding module and encoding module of the preset deep learning model may include: performing at least one wavelet analysis in the decoding module and encoding module of the preset deep learning model to obtain at least one wavelet decomposition two-dimensional sequence; performing two-dimensional feature extraction and information restoration on the at least one wavelet decomposition two-dimensional sequence, and performing Fourier aggregation analysis on the sample data of the queuing passenger flow at the outlets.

[0058] The wavelet decomposition two-dimensional sequence can be a two-dimensional time series obtained by wavelet decomposition of sample data of customer flow in queues at outlets. Fourier aggregation analysis can be used to extract frequency domain features from time domain or spatial domain data.

[0059] In an embodiment of the present invention, the decoding module and the encoding module of the preset deep learning model can be subjected to at least one wavelet analysis respectively to obtain at least one wavelet decomposition two-dimensional sequence, thereby utilizing the decomposition convolution mechanism to extract two-dimensional features from the wavelet decomposition two-dimensional sequence, further restoring information on the extracted two-dimensional features, and performing Fourier aggregation analysis on the sample data of the queuing passenger flow at the outlets.

[0060] In an optional embodiment of the present invention, performing two-dimensional feature extraction and information restoration on at least one wavelet decomposition two-dimensional sequence may include: performing two-dimensional feature extraction on the wavelet decomposition two-dimensional sequence based on convolution pooling modules of different sizes in a preset deep learning model to obtain target two-dimensional extracted features; filtering the target two-dimensional extracted features to obtain one-dimensional filtered features, and performing information restoration on the one-dimensional filtered features based on a long short-term memory network in the preset deep learning model.

[0061] The target two-dimensional extracted features may be features between cycles and within cycles extracted from a two-dimensional wavelet decomposition sequence, and the one-dimensional filtered features may be one-dimensional features obtained by filtering the target two-dimensional extracted features.

[0062] In an embodiment of the present invention, convolution pooling modules of different sizes are set in the preset deep learning model to perform two-dimensional feature extraction on the wavelet decomposition two-dimensional sequence through the convolution pooling modules of different sizes to obtain the target two-dimensional extracted features, and the target two-dimensional extracted features are filtered using the filters set in the preset deep learning model to obtain one-dimensional filtered features, thereby restoring information of the one-dimensional filtered features based on the long short-term memory network in the preset deep learning model.

[0063] In an optional embodiment of the present invention, the preset deep learning model may include a feedforward network, a first wavelet filter, and a second wavelet filter.

[0064] The first wavelet filter may be a high-frequency filter, and the second wavelet filter may be a low-frequency filter.

[0065] In an optional embodiment of the present invention, after inputting the sample data of the outlet queuing passenger flow into a preset deep learning model, performing serial time change feature analysis and time dependency analysis between data, and obtaining the outlet queuing passenger flow prediction model, it may also include: determining the prediction error analysis index of the outlet queuing passenger flow prediction model; wherein the prediction error analysis index may include mean square error, and / or mean absolute error; based on the prediction error analysis index, performing parameter optimization on the outlet queuing passenger flow prediction model.

[0066] Among them, the prediction error analysis indicator can be used to describe the magnitude of the prediction error of the network queuing customer flow prediction model. The mean squared error can be the mean of the sum of the squares of the errors between the observed values (actual values) and the predicted values. The mean absolute error can be the mean of the absolute values of the differences between the actual values and the predicted values.

[0067] In an embodiment of the present invention, the outlet queuing passenger flow prediction model can be verified using a test set and a validation set corresponding to the outlet queuing passenger flow sample data, and the mean square error and / or mean absolute error can be calculated. The calculated errors are used as prediction error analysis indicators, and then the parameters of the outlet queuing passenger flow prediction model are optimized based on the difference between the model error allowable error and the prediction error analysis indicator, so that the optimized outlet queuing passenger flow prediction model meets the model error allowable error.

[0068] In an optional embodiment of the present invention, the method for predicting customer flow in a network queue may further include: obtaining data related to customer flow prediction in a network queue; and analyzing the data related to customer flow prediction in a network queue using a model for predicting customer flow in a network queue to obtain an estimated customer flow in a network queue. The model for predicting customer flow in a network queue is a model trained using the model training method of any of the embodiments.

[0069] The branch queue customer flow prediction-related data may be data that needs to be input into the branch queue customer flow prediction model to perform branch queue customer flow prediction. The branch queue estimated customer flow may be the analysis result of the branch queue customer flow prediction-related data by the branch queue customer flow prediction model.

[0070] In an embodiment of the present invention, the outlet queue passenger flow prediction associated data of the current business outlets can be obtained in real time, and then the outlet queue passenger flow prediction associated data can be input into the outlet queue passenger flow prediction model to analyze the outlet queue passenger flow prediction associated data through the outlet queue passenger flow prediction model to obtain the outlet queue estimated passenger flow.

[0071] The technical solution of the embodiment of the present invention obtains historical data of the branch queue passenger flow of at least one business branch, thereby cleaning and preprocessing the historical data of the branch queue passenger flow to obtain branch queue passenger flow sample data, and then inputs the branch queue passenger flow sample data into a preset deep learning model to perform sequence time change feature analysis and time dependency analysis between data to obtain a branch queue passenger flow prediction model. Since the modeling of actual queue passenger flow involves multiple variables, the sequence time change feature analysis and training of the preset deep learning model based on the branch queue passenger flow sample data can consider the interaction between multiple variables and their correlation characteristics with events to learn the periodicity between sequences and the time changes between cycles. The analysis and training of time dependency between data can improve the ability to capture the dependency between data. Compared with the traditional neural network model, it can also effectively reduce the computational complexity, solve the problems of poor prediction accuracy and efficiency of existing queue passenger flow prediction methods, and can improve the prediction accuracy and efficiency of queue passenger flow.

[0072] Example 3

[0073] The third embodiment of the present invention provides an optional embodiment of a method for predicting customer flow in a network station queue, and its specific implementation can be found in the following embodiments.

[0074] This solution provides a technical solution based on improved Transformer, such as Figure 3 As shown in the figure, the wavelet decomposition in the two-dimensional feature module is first used to convert the one-dimensional network queue passenger flow prediction correlation data into a two-dimensional tensor, and the complex periodic features in the two-dimensional tensor are captured through Fourier aggregation analysis. Then, the autoformer with autocorrelation mechanism, the first filter and the second filter are used to predict the queue passenger flow at the next moment. Figure 3 In the equation, ① represents the two-dimensional feature module, ② represents the feedforward, A represents the Autoformer of the autocorrelation mechanism, B1 represents the first wavelet filter, and B2 represents the second wavelet filter. Assume that the input of the two-dimensional feature module is The queuing passenger flow prediction result is

[0075] The two-dimensional feature module is used to mine the periodic characteristics of the sequence at a deeper level. The two-dimensional feature model includes a wavelet decomposition module and a time feature acquisition module. The time feature acquisition module includes a three-layer convolution pooling module and a long short-term memory network. The convolution pooling module uses the Inception (initial network) + FFT-Softmax (Fourier aggregation) layer to gradually extract the intra-period and inter-cycle feature information of the time series data. Specifically, convolution kernels of different sizes of 1×1 and 3×3 are used to obtain different receptive fields and downsample to learn information between different cycles. The three-layer convolution pooling module and the two-dimensional feature module are connected through the FFT-Softmax layer residual connection, and then the detailed information is restored through the long short-term memory network to further learn long-term time series dependencies. Its overall architecture is as follows: Figure 4 shown.

[0076] By inserting a two-dimensional feature module into the encoder and decoder architecture, the original model's autoregressive module is retained, further exploring the periodic characteristics of the sequence in the time dimension. Because the model uses a wavelet decomposition module to transform the one-dimensional time series into a two-dimensional space, various image processing convolutional networks can be used to learn the intra-cycle and inter-cycle features of the two-dimensional time series.

[0077] Wavelet decomposition is used to completely separate high-frequency and low-frequency information in the data, and feature extraction is used to learn the important periodicity and trend characteristics of the sequence. Because multivariate long time series typically exhibit large volatility and temporal continuity, an orthogonal wavelet is selected to perform an N-order decomposition of the data (denoted as dbN). The dbN wavelet has a nonlinear phase and no fixed kernel function, resulting in good regularization. The signal reconstruction process is relatively smooth, and it has strong frequency domain localization and frequency band partitioning effects, clearly decomposing the data into multiple periodic and trend sequences.

[0078] The two-dimensional feature module adopts the improved InceptionV3 architecture and combines the decomposition convolution idea. Since the existing model is mainly used for image processing tasks with large depth and width, the size of its convolution kernel cannot adapt to the smaller periodic decomposition level in this solution. Figure 5 As shown in the figure, the Inception network used contains 4 convolution branches, and each layer in the network can learn the features of "multi-cycle" (3×3, 5×5) or "adjacent cycle" (1×1).

[0079] This scheme integrates long short-term memory networks into the embedding layer to obtain more powerful long-term history representation while mapping the input data.

[0080] Bank customer flow data is affected by factors such as weather and holidays, exhibiting certain periodicities and trends. Wavelet decomposition and two-dimensional convolutional neural networks can effectively capture the complex periodic characteristics of multivariate time series. The traditional Transformer model uses a point-by-point dot product attention mechanism, which has excessively high computational complexity. The improved Transformer model, which uses Fourier transform aggregation connections, not only significantly reduces the computational complexity but also improves its ability to capture the long-term dependencies of customer flow data. Traditional prediction models use single-variable prediction methods, while this solution adopts a multivariate prediction method to achieve multiple inputs and multiple outputs, avoiding the difficulty of collecting influencing variables during the prediction phase.

[0081] Example 4

[0082] Figure 6 This is a structural diagram of a model training device provided by the fourth embodiment of the present invention. Figure 6 As shown, the device includes:

[0083] The sample data acquisition module 310 is used to obtain sample data of customer flow in queues at outlets;

[0084] Model training model 320, for inputting sample data of customer flow in queues at outlets into a preset deep learning model, performing analysis on the time variation characteristics of the sequence and the time dependency between the data, and obtaining a customer flow prediction model for queues at outlets;

[0085] Among them, the outlet queue passenger flow prediction model is used to predict the outlet queue passenger flow.

[0086] The technical solution of an embodiment of the present invention obtains sample data of customer flow in queues at outlets, inputs the sample data into a preset deep learning model, performs analysis of the time-varying characteristics of the sequence and the temporal dependency between data, and obtains an outlet queue flow prediction model. Because the modeling of actual queue flow involves multiple variables, the training of the preset deep learning model for the analysis of the time-varying characteristics of the sequence based on the sample data of customer flow in queues at outlets can consider the interactions between multiple variables and their associated characteristics with events to learn the periodicity between sequences and the temporal variation between cycles. The training of the analysis of the temporal dependency between data can improve the ability to capture the dependencies between data. Compared with traditional neural network models, it can also effectively reduce the computational complexity, solve the problems of poor prediction accuracy and efficiency of existing queue flow prediction methods, and can improve the prediction accuracy and efficiency of queue flow.

[0087] Optionally, the sample data acquisition module 310 is specifically configured to acquire historical data of customer flow in queues at at least one business outlet; and clean and pre-process the historical data of customer flow in queues at the outlets to obtain sample data of customer flow in queues at the outlets.

[0088] Optionally, the model training model 320 is specifically used to input the sample data of the queuing passenger flow at the outlet into a preset deep learning model, and perform at least one sequence time change feature analysis and time dependency analysis between data in the decoding module and encoding module of the preset deep learning model to obtain the queuing passenger flow prediction model for the outlet.

[0089] Optionally, the model training model 320 is specifically used to perform at least one wavelet analysis in the decoding module and the encoding module of the preset deep learning model to obtain at least one wavelet decomposition two-dimensional sequence; perform two-dimensional feature extraction and information restoration on the at least one wavelet decomposition two-dimensional sequence, and perform Fourier aggregation analysis on the sample data of the queuing passenger flow at the outlet.

[0090] Optionally, the model training model 320 is specifically used to perform two-dimensional feature extraction on the wavelet decomposition two-dimensional sequence based on convolution pooling modules of different sizes in the preset deep learning model to obtain target two-dimensional extracted features; perform filtering processing on the target two-dimensional extracted features to obtain one-dimensional filtered features, and perform information restoration on the one-dimensional filtered features based on the long short-term memory network in the preset deep learning model.

[0091] Optionally, the preset deep learning model includes a feedforward network, a first wavelet filter, and a second wavelet filter.

[0092] Optionally, the model training device also includes a model optimization module, which is used to determine the prediction error analysis index of the network queuing passenger flow prediction model; wherein the prediction error analysis index includes the mean square error and / or the mean absolute error; based on the prediction error analysis index, the parameters of the network queuing passenger flow prediction model are optimized.

[0093] The model training device provided in the embodiment of the present invention can execute the model training method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0094] An embodiment of the present invention also provides a network point queuing passenger flow prediction device, which includes a real-time data acquisition module and a passenger flow prediction module.

[0095] Real-time data acquisition module, used to obtain data related to the prediction of customer flow in queues at outlets;

[0096] The passenger flow prediction module is used to analyze the network queue passenger flow prediction related data through the network queue passenger flow prediction model to obtain the estimated network queue passenger flow; wherein, the network queue passenger flow prediction model is a model trained by the model training method in any embodiment.

[0097] The network point queuing passenger flow prediction device provided by the embodiment of the present invention can execute the network point queuing passenger flow prediction method provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0098] Example 5

[0099] Figure 7 A schematic diagram of an electronic device that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The components shown herein, their connections and relationships, and their functions are provided for example only and are not intended to limit the implementation of the present inventions described and / or claimed herein.

[0100] like Figure 7 As shown, electronic device 10 includes at least one processor 11 and memory, such as ROM 12 and RAM 13, communicatively connected to at least one processor 11. The memory stores computer programs executable by the at least one processor, and processor 11 can perform various appropriate actions and processes based on the computer programs stored in ROM 12 or loaded from storage unit 18 into RAM 13. RAM 13 can also store various programs and data required for the operation of electronic device 10. Processor 11, ROM 12, and RAM 13 are interconnected via bus 14. An I / O interface 15 is also connected to bus 14. ROM 12 is a read-only memory, RAM 13 is a random access memory, and I / O interface 15 is an input / output interface.

[0101] Multiple components in the electronic device 10 are connected to the I / O interface 15, including an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0102] Processor 11 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various processors that run machine learning model algorithms, digital signal processors (DSPs), and any other suitable processors, controllers, microcontrollers, etc. Processor 11 executes the various methods and processes described above, such as the model training method or the method for predicting customer flow in a network queue.

[0103] In some embodiments, the model training method or the method for predicting the number of passengers queuing at an outlet may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the model training method or the method for predicting the number of passengers queuing at an outlet described above may be executed. Alternatively, in other embodiments, the processor 11 may be configured to execute the model training method or the method for predicting the number of passengers queuing at an outlet by any other appropriate means (e.g., by means of firmware).

[0104] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0105] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the computer program is executed by the processor, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer program may be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0106] In the context of the present invention, computer-readable storage medium can be a tangible medium that can contain or store a computer program for use with an instruction execution system, device or equipment or used in combination with an instruction execution system, device or equipment. Computer-readable storage medium can include but is not limited to electronic, magnetic, optical, electromagnetic, infrared or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, computer-readable storage medium can be a machine-readable signal medium. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, RAM, ROM, an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device or any suitable combination of the foregoing.

[0107] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0108] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0109] A computing system may include clients and servers. The clients and servers are generally remote from each other and typically interact via a communication network. This client-server relationship arises through computer programs running on the respective computers, creating a client-server relationship. The server may be a cloud server, also known as a cloud computing server or cloud host. This server is a hosting product within the cloud computing service ecosystem that addresses the management difficulties and limited scalability of traditional physical hosts and VPS servers.

[0110] The present application also discloses a computer program product comprising a computer program that, when executed by a processor, implements the model training method or the method for predicting customer flow in a network queue provided in any of the embodiments of the present application. This program product embodies the same inventive concept as the model training method or the method for predicting customer flow in a network queue disclosed in each embodiment of the present application, and therefore is not further described here.

[0111] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present invention can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved. This is not limited herein.

[0112] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A model training method, characterized in that: include: Obtain sample data on customer flow in queues at outlets; Inputting the sample data of the customer flow in queue at the outlet into a preset deep learning model, performing sequence time variation feature analysis and time dependency analysis between the data, and obtaining a customer flow prediction model for the outlet queue; The network point queue passenger flow prediction model is used to predict the network point queue passenger flow.

2. The model training method according to claim 1, characterized in that Obtain sample data on customer queuing at outlets, including: Obtain historical customer flow data of at least one business outlet; The historical data of the customer flow in queue at the outlet is cleaned and preprocessed to obtain sample data of the customer flow in queue at the outlet.

3. The model training method according to claim 1, characterized in that The sample data of the customer flow in queues at the outlets are input into a preset deep learning model to perform sequence time variation feature analysis and time dependency analysis between data to obtain an outlet queue customer flow prediction model, including: The sample data of the queuing passenger flow at the outlet is input into a preset deep learning model, and at least one sequence time change feature analysis and time dependency analysis between data are performed in the decoding module and encoding module of the preset deep learning model to obtain the queuing passenger flow prediction model for the outlet.

4. The model training method according to claim 3, characterized in that In the decoding module and the encoding module of the preset deep learning model, at least one sequence time change feature analysis is performed, including: Performing at least one wavelet analysis in the decoding module and the encoding module of the preset deep learning model to obtain at least one wavelet decomposition two-dimensional sequence; Two-dimensional feature extraction and information restoration are performed on the at least one wavelet decomposition two-dimensional sequence, and Fourier aggregation analysis is performed on the sample data of the customer flow in queues at the outlets.

5. The model training method according to claim 4, characterized in that Performing two-dimensional feature extraction and information restoration on the at least one wavelet decomposition two-dimensional sequence includes: Based on the convolution pooling modules of different sizes in the preset deep learning model, two-dimensional feature extraction is performed on the wavelet decomposition two-dimensional sequence to obtain target two-dimensional extracted features; The target two-dimensional extracted features are filtered to obtain one-dimensional filtered features, and information restoration is performed on the one-dimensional filtered features based on the long short-term memory network in the preset deep learning model.

6. The model training method according to claim 1, characterized in that The preset deep learning model includes a feedforward network, a first wavelet filter and a second wavelet filter.

7. The model training method according to claim 1, characterized in that After inputting the sample data of the customer flow in queue at the outlet into a preset deep learning model, performing a sequence time variation feature analysis and a time dependency analysis between the data, and obtaining an outlet customer flow prediction model, the method further includes: Determining a prediction error analysis index of the network point queuing passenger flow prediction model; wherein the prediction error analysis index includes a mean square error and / or a mean absolute error; Based on the prediction error analysis index, the parameters of the network point queuing passenger flow prediction model are optimized.

8. A method for predicting customer flow in queues at a network point, characterized in that: include: Obtaining data related to customer flow prediction at outlets; Analyze the customer flow prediction data of the network queues by using the network queue customer flow prediction model to obtain the estimated customer flow of the network queues; Wherein, the network queuing passenger flow prediction model is a model trained by the model training method described in any one of claims 1-7.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the model training method described in any one of claims 1 to 7, or execute the network queue passenger flow prediction method described in claim 8.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the model training method described in any one of claims 1 to 7 when executed, or to execute the method for predicting customer flow in a queue at a network point described in claim 8.

Images