Method and device for predicting aviation flow

Abstract

The invention discloses an aviation flow prediction method and device, and relates to the technical field of data processing, historical aviation flow is subjected to multi-resolution time sequence feature coding, a multi-resolution time sequence feature coding sequence is generated, the multi-resolution time sequence feature coding sequence comprises significant multi-resolution time sequence features corresponding to the historical aviation flow, and meanwhile, the significant multi-resolution time sequence features correspond to the historical aviation flow. According to the method, a multi-resolution time sequence feature coding sequence is subjected to data splicing processing with a time-invariant feature and a time-varying feature, time sequence feature enhancement of the time-varying feature and the time-invariant feature is achieved, and therefore an aviation flow prediction model can capture time sequence association on aviation flow changes more easily; the normalized historical aviation flow is decomposed based on Fourier transform to obtain a time-invariant sequence and a time-varying sequence, so that the perception of the aviation flow prediction model on different time sequence frequencies is realized; by integrating the mechanism into the aviation flow prediction model, the performance breakthrough of the aviation flow prediction model and the improvement of the accuracy of the prediction result are realized.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to a method and device for predicting aviation traffic. Background Art

[0002] With the continuous development of the aviation industry, the increase in air traffic has put forward higher requirements for air traffic management. Air Traffic Flow Forecasting (ATFP) technology has played a key role in this context, providing important support for improving airspace management efficiency and optimizing flight scheduling. Accurate ATFP can help air traffic controllers predict the flow conditions of the airspace in advance, and then implement more efficient traffic management measures to avoid congestion, reduce delays, improve the utilization efficiency of airspace resources, and ensure the smooth and safe operation of flights.

[0003] The existing ATFP method mainly focuses on the characteristics of air traffic changes reflected in each air traffic data series itself, which is specifically manifested in directly modeling each input air traffic data series to capture the law of air traffic changes. However, directly modeling the air traffic data series itself often fails to capture the complex connections between the changes in each air traffic, resulting in information loss and decreased accuracy. Moreover, the existing ATFP method often relies on the prediction model itself to implicitly mine the time series characteristics of air traffic, which is in a "black box state". The time series characteristics in the "black box state" are often unable to be fully utilized by the prediction model, which leads to the existing prediction model always having performance bottlenecks, low accuracy of air traffic prediction results and insufficient interpretability.

[0004] Therefore, it is an urgent problem to provide a method and device for predicting air traffic so that the existing prediction model can break through the performance bottleneck and have time series interpretability, and improve the accuracy of air traffic prediction in complex airspace scenarios. Summary of the invention

[0005] The present invention provides a method and device for predicting air traffic flow, the main purpose of which is to provide a method and device for predicting air traffic flow, so as to enable the existing prediction model to break through the performance bottleneck and improve the accuracy of air traffic flow prediction in complex airspace scenarios.

[0006] According to a first aspect of the present disclosure, a method for predicting aviation traffic is provided, comprising:

[0007] Receive air traffic data and pre-process it to obtain historical air traffic;

[0008] Performing multi-resolution time series feature encoding on the historical air traffic to generate a multi-resolution time series feature encoding sequence;

[0009] Decomposing the normalized historical air traffic based on Fourier transform to obtain a time-invariant sequence and a time-varying sequence;

[0010] Extracting features from the time-invariant sequence and the time-varying sequence respectively to obtain time-invariant features and time-varying features;

[0011] The multi-resolution time series feature coding sequence is respectively subjected to data splicing processing with the time-invariant feature and the time-varying feature to obtain time series enhanced time-varying sequence embedded data and time series enhanced time-invariant sequence embedded data;

[0012] Based on the long short-term memory network, the time-varying sequence embedding data and the time-invariant sequence embedding data are learned respectively to obtain the time-invariant time series features and the time-varying time series features;

[0013] Based on the attention scores corresponding to the time-invariant temporal features and the time-varying temporal features, weighted processing is performed respectively to obtain a weighted time-invariant temporal feature sequence and a weighted time-varying temporal feature;

[0014] Based on the KAN prediction network, the air traffic flow is predicted for the weighted time-invariant time series feature sequence and the weighted time-varying time series feature, and the predicted time-invariant air traffic flow sequence and the predicted time-varying air traffic flow sequence are obtained.

[0015] The sum of the predicted time-invariant aviation traffic sequence and the predicted time-varying aviation traffic sequence is subjected to denormalization processing to obtain an actual aviation traffic prediction result.

[0016] Optionally, performing multi-resolution time series feature encoding on the historical air traffic to generate a multi-resolution time series feature encoding sequence includes:

[0017] Performing time series feature encoding with time resolution on the historical air traffic to obtain a time series feature encoding sequence;

[0018] Performing time series feature encoding with hourly resolution on the historical air traffic to obtain an hourly time series feature encoding sequence;

[0019] The historical air traffic is encoded with time series features at a weekly resolution to obtain a weekly time series feature coding sequence, wherein the multi-resolution time series feature coding sequence includes the moment time series feature coding sequence, the hour time series feature coding sequence and the weekly time series feature coding sequence.

[0020] Optionally, the decomposing the normalized historical air traffic based on Fourier transform to obtain a time-invariant sequence and a time-varying sequence includes:

[0021] Transforming the normalized time domain sequence of the historical air traffic into a frequency domain complex sequence based on discrete Fourier transform;

[0022] Filtering the high-frequency components of the frequency-domain complex sequence based on a low-pass filter, and converting the filtering result from the frequency domain to the time domain to obtain the time-invariant sequence;

[0023] The difference between the time domain sequence and the time invariant sequence is calculated to obtain the time varying sequence.

[0024] Optionally, the extracting features from the time-invariant sequence and the time-varying sequence respectively to obtain the time-invariant features and the time-varying features comprises:

[0025] Performing linear mapping processing on the time-invariant sequence and the time-varying sequence respectively to obtain an aligned time-invariant sequence aligned to a first preset sequence length and an aligned time-varying sequence aligned to a second preset sequence length;

[0026] Based on the convolution layer, high-dimensional features are extracted for the aligned time-invariant sequence and the aligned time-varying sequence respectively to obtain the time-invariant features and the time-varying features.

[0027] Optionally, the step of performing data splicing processing on the multi-resolution time series feature coding sequence with the time-invariant feature and the time-varying feature to obtain time series enhanced time-varying sequence embedded data and time series enhanced time-invariant sequence embedded data comprises:

[0028] Performing high-dimensional feature extraction on the moment time series feature coding sequence, the hour time series feature coding sequence, and the week time series feature coding sequence, and concatenating the three feature extraction results to obtain a potential time series representation;

[0029] Concatenating the latent temporal representation with the time-invariant feature to obtain the temporal enhanced time-invariant sequence embedding data;

[0030] The latent temporal representation and the time-varying feature are concatenated to obtain the temporal enhanced time-varying sequence embedding data.

[0031] Optionally, the receiving and preprocessing of air traffic data to obtain historical air traffic includes:

[0032] Converting the format of the air traffic data acquired and updated in real time into a structured data format suitable for subsequent processing;

[0033] Extract key data fields from the air traffic data after format conversion and remove duplicate data;

[0034] A time series is constructed based on the deduplicated key data fields according to the historical time step length to generate the historical aviation traffic.

[0035] Optionally, before receiving and preprocessing the air traffic data to obtain historical air traffic, the method includes:

[0036] Training a preset deep learning framework to obtain an air traffic prediction model, wherein the air traffic prediction model is used to execute the air traffic prediction method;

[0037] The aviation traffic prediction model includes a frequency perception module, a sequence modeling module and a KAN prediction module;

[0038] The frequency perception module is used to decompose and extract features of the historical air traffic input into the air traffic prediction model to generate time series features, wherein the time series features include the time-invariant features and the time-varying features;

[0039] The sequence modeling module for enhancing the temporal features is used to fuse the temporal features with the multi-resolution temporal feature encoding sequence, and extract temporal dependency from the fusion result based on the long short-term memory network;

[0040] The KAN prediction module based on the attention mechanism is used to map the weighted time-invariant temporal feature sequence and the weighted time-varying temporal feature to prediction dimensions respectively and output a predicted air traffic sequence, wherein the predicted air traffic sequence includes the predicted time-invariant air traffic sequence and the predicted time-varying air traffic sequence.

[0041] Optionally, the training of a preset deep learning framework to obtain an air traffic prediction model includes:

[0042] The constructed training set is used to train the air traffic prediction model, and the loss function is used for back propagation to update the network parameters of the air traffic prediction model. The Adam optimizer is used to dynamically adjust the learning rate. The formula of the loss function is expressed as follows:

[0043] L dd =L time +αL freq

[0044] Among them, L dd is the loss function, L time is the time domain loss, L freq is the frequency domain loss, α represents the scaling factor, which is used to control the impact of the frequency domain loss on the total loss;

[0045] The calculation formula of the time domain loss is as follows:

[0046]

[0047] Among them, x ij represents the true value of the time domain series, represents the predicted value of the time domain sequence, T represents the length of the predicted time domain sequence, N is the number of time domain sequence samples in a training batch, i and j represent subscript variables, i is the length index, and j is the sample index;

[0048] The frequency domain loss is calculated as follows:

[0049] y=DFT(x),

[0050]

[0051] Where DFT(·) represents discrete Fourier transform, x is the time domain sequence, y is the frequency domain sequence, To predict the time domain series, To predict the frequency domain sequence, y ij is the true value of the frequency domain sequence, is the predicted value of the frequency domain series, represents the length of the frequency domain sequence after the real Fourier transform, T represents the length of the predicted time domain sequence before the real Fourier transform, n is the number of frequency domain sequence samples in a training batch, i and j represent subscript variables, i is the length index, and j is the sample index.

[0052] According to a second aspect of the present disclosure, there is provided an aviation traffic prediction device, comprising:

[0053] A receiving unit, used for receiving and preprocessing the air traffic data to obtain the historical air traffic;

[0054] An encoding unit, used for performing multi-resolution time series feature encoding on the historical air traffic to generate a multi-resolution time series feature encoding sequence;

[0055] A processing unit, configured to decompose the normalized historical air traffic based on Fourier transform to obtain a time-invariant sequence and a time-varying sequence;

[0056] An extraction unit, used to extract features from the time-invariant sequence and the time-varying sequence respectively to obtain a time-invariant feature and a time-varying feature;

[0057] A splicing unit, used for performing data splicing processing on the multi-resolution time series feature coding sequence with the time-invariant feature and the time-varying feature respectively, to obtain time series enhanced time-varying sequence embedded data and time series enhanced time-invariant sequence embedded data;

[0058] A learning unit, used for learning the time-varying sequence embedding data and the time-invariant sequence embedding data based on a long short-term memory network, respectively, to obtain a time-invariant time series feature and a time-varying time series feature;

[0059] A weighting unit, configured to perform weighted processing based on the attention scores corresponding to the time-invariant temporal features and the time-varying temporal features, respectively, to obtain a weighted time-invariant temporal feature sequence and a weighted time-varying temporal feature;

[0060] A prediction unit is used to perform air traffic prediction on the weighted time-invariant time series feature sequence and the weighted time-varying time series feature based on a KAN prediction network, to obtain a predicted time-invariant air traffic sequence and a predicted time-varying air traffic sequence;

[0061] The denormalization unit is used to perform denormalization processing on the sum of the predicted time-invariant aviation traffic sequence and the predicted time-varying aviation traffic sequence to obtain the actual aviation traffic prediction result.

[0062] According to a third aspect of the present disclosure, there is provided an electronic device, including:

[0063] at least one processor; and

[0064] a memory communicatively connected to the at least one processor; wherein,

[0065] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method described in the first aspect.

[0066] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute the method described in the first aspect.

[0067] According to a fifth aspect of the present disclosure, a computer program product is provided, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements the method as described in the first aspect above.

[0068] The present invention provides a method and device for predicting air traffic flow, which receives and preprocesses air traffic flow data to obtain historical air traffic flow; performs multi-resolution time series feature encoding on the historical air traffic flow to generate a multi-resolution time series feature encoding sequence; decomposes the normalized historical air traffic flow based on Fourier transform to obtain a time-invariant sequence and a time-varying sequence; extracts features from the time-invariant sequence and the time-varying sequence to obtain time-invariant features and time-varying features; performs data splicing processing on the multi-resolution time series feature encoding sequence with the time-invariant features and the time-varying features to obtain time-series enhanced time-varying sequence embedded data and time-series enhanced time-invariant sequence embedded data; based on length and short The period memory network learns the time-varying sequence embedding data and the time-invariant sequence embedding data respectively to obtain the time-invariant time series features and the time-varying time series features; based on the attention scores corresponding to the time-invariant time series features and the time-varying time series features, weighted processing is performed respectively to obtain the weighted time-invariant time series feature sequence and the weighted time-varying time series feature; based on the KAN prediction network, the weighted time-invariant time series feature sequence and the weighted time-varying time series feature are used to predict air traffic flow respectively to obtain the predicted time-invariant air traffic flow sequence and the predicted time-varying air traffic flow sequence; the sum of the predicted time-invariant air traffic flow sequence and the predicted time-varying air traffic flow sequence is denormalized to obtain the actual air traffic prediction result. Compared with the related art, the multi-resolution time series feature coding sequence is generated by performing multi-resolution time series feature coding on the historical air traffic, and the multi-resolution time series feature coding sequence includes significant multi-resolution time series features corresponding to the historical air traffic. At the same time, the multi-resolution time series feature coding sequence is respectively subjected to data splicing processing with the time-invariant feature and the time-varying feature to achieve time series feature enhancement of the time-varying feature and the time-invariant feature, so that the air traffic prediction model can more easily capture the time series correlation of air traffic changes. In addition, the normalized historical air traffic is decomposed based on Fourier transform to obtain time-invariant sequence and time-varying sequence, so as to realize the perception of the air traffic prediction model to different time series frequencies. Furthermore, by integrating the aforementioned mechanism into the air traffic prediction model, the performance breakthrough of the air traffic prediction model and the improvement of the accuracy of the prediction results are achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0070] Figure 1 A schematic diagram of a flow chart of a method for predicting air traffic flow provided by an embodiment of the present disclosure;

[0071] Figure 2A schematic diagram of the overall structure of an air traffic prediction model provided by an embodiment of the present disclosure;

[0072] Figure 3 A schematic diagram of the structure of a decomposition module provided by an embodiment of the present disclosure;

[0073] Figure 4 A schematic diagram of the structure of a KAN prediction module provided by an embodiment of the present disclosure;

[0074] Figure 5 A schematic diagram of the structure of an aviation traffic prediction device provided by an embodiment of the present disclosure;

[0075] Figure 6 A schematic block diagram of an exemplary electronic device 300 provided for an embodiment of the present disclosure. DETAILED DESCRIPTION

[0076] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present invention can be combined with each other without conflict.

[0077] The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise specified, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art to which the present invention belongs. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present invention.

[0078] In addition, the terms "first", "second", etc. in the specification and claims of the present disclosure 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 terms used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in sequences other than those illustrated or described herein.

[0079] The following describes the aviation traffic prediction method and device according to the embodiments of the present disclosure with reference to the accompanying drawings.

[0080] In order to at least provide a method and device for predicting air traffic, so as to enable the existing prediction model to break through the performance bottleneck and improve the accuracy of air traffic prediction in complex airspace scenarios, this embodiment provides a method for predicting air traffic.

[0081] Figure 1 The following is a flow chart of a method for predicting air traffic flow provided by an embodiment of the present disclosure. Figure 1 As shown, the method comprises the following steps:

[0082] Step 101, receiving air traffic data and preprocessing it to obtain historical air traffic;

[0083] Step 102, performing multi-resolution time series feature encoding on the historical air traffic to generate a multi-resolution time series feature encoding sequence;

[0084] Step 103, decomposing the normalized historical air traffic based on Fourier transform to obtain a time-invariant sequence and a time-varying sequence;

[0085] Step 104, extracting features from the time-invariant sequence and the time-varying sequence respectively to obtain time-invariant features and time-varying features;

[0086] Step 105, performing data splicing processing on the multi-resolution time series feature coding sequence, the time-invariant feature and the time-varying feature respectively, to obtain time series enhanced time-varying sequence embedded data and time series enhanced time-invariant sequence embedded data;

[0087] Step 106, learning the time-varying sequence embedding data and the time-invariant sequence embedding data based on the long short-term memory network respectively to obtain the time-invariant time series features and the time-varying time series features;

[0088] Step 107, performing weighted processing based on the attention scores corresponding to the time-invariant temporal features and the time-varying temporal features, respectively, to obtain a weighted time-invariant temporal feature sequence and a weighted time-varying temporal feature;

[0089] Step 108, based on the KAN prediction network, perform air traffic prediction on the weighted time-invariant time series feature sequence and the weighted time-varying time series feature to obtain a predicted time-invariant air traffic sequence and a predicted time-varying air traffic sequence;

[0090] Step 109, performing a denormalization process on the sum of the predicted time-invariant air traffic flow sequence and the predicted time-varying air traffic flow sequence to obtain an actual air traffic flow prediction result.

[0091] The present invention provides an air traffic prediction method, which receives air traffic data and performs preprocessing to obtain historical air traffic; performs multi-resolution time series feature encoding on the historical air traffic to generate a multi-resolution time series feature encoding sequence; decomposes the normalized historical air traffic based on Fourier transform to obtain a time-invariant sequence and a time-varying sequence; extracts features from the time-invariant sequence and the time-varying sequence to obtain time-invariant features and time-varying features; performs data splicing processing on the multi-resolution time series feature encoding sequence with the time-invariant features and the time-varying features to obtain time-series enhanced time-varying sequence embedded data and time-series enhanced time-invariant sequence embedded data; based on long-term and short-term The memory network learns the time-varying sequence embedding data and the time-invariant sequence embedding data respectively to obtain the time-invariant time series features and the time-varying time series features; weighted processing is performed on the attention scores corresponding to the time-invariant time series features and the time-varying time series features respectively to obtain the weighted time-invariant time series feature sequence and the weighted time-varying time series feature; based on the KAN prediction network, the weighted time-invariant time series feature sequence and the weighted time-varying time series feature are used to predict air traffic flow to obtain the predicted time-invariant air traffic flow sequence and the predicted time-varying air traffic flow sequence; the sum of the predicted time-invariant air traffic flow sequence and the predicted time-varying air traffic flow sequence is denormalized to obtain the actual air traffic prediction result. Compared with the related art, the multi-resolution time series feature coding sequence is generated by performing multi-resolution time series feature coding on the historical air traffic, and the multi-resolution time series feature coding sequence includes significant multi-resolution time series features corresponding to the historical air traffic. At the same time, the multi-resolution time series feature coding sequence is respectively subjected to data splicing processing with the time-invariant feature and the time-varying feature to achieve time series feature enhancement of the time-varying feature and the time-invariant feature, so that the air traffic prediction model can more easily capture the time series correlation of air traffic changes. In addition, the normalized historical air traffic is decomposed based on Fourier transform to obtain time-invariant sequence and time-varying sequence, so as to realize the perception of the air traffic prediction model to different time series frequencies. Furthermore, by integrating the aforementioned mechanism into the air traffic prediction model, the performance breakthrough of the air traffic prediction model and the improvement of the accuracy of the prediction results are achieved.

[0092] As a refinement of the embodiment of the present disclosure, when executing step 102 to perform multi-resolution time series feature encoding on the historical air traffic and generate a multi-resolution time series feature encoding sequence, the following implementation methods may also be used but are not limited to, for example: performing time series feature encoding on the historical air traffic with moment resolution to obtain a moment time series feature encoding sequence; performing time series feature encoding on the historical air traffic with hour resolution to obtain an hour time series feature encoding sequence; performing time series feature encoding on the historical air traffic with week resolution to obtain a week time series feature encoding sequence, and the multi-resolution time series feature encoding sequence includes the moment time series feature encoding sequence, the hour time series feature encoding sequence, and the week time series feature encoding sequence.

[0093] As an extension of the above embodiment, the time series feature coding sequence (Quarter-Of-Day, QoD) is used to capture the local fluctuation of air traffic at each time of the day, and its calculation process is shown in formula (1):

[0094]

[0095] Among them, ts represents the timestamp corresponding to the current time step, and QoD(·) represents the function that encodes the calculation time of ts. In the function expression, ts.minute and ts.hour represent the minute and hour corresponding to the timestamp respectively, and finally minus one to convert the code into a base 0 index.

[0096] The hourly time series feature coding sequence is used to assist the aviation traffic prediction model in capturing the air traffic characteristics at the hourly scale within a day, and its calculation formula is:

[0097] HoD(ts)=ts.hour-1,whereHoD(ts)∈{0,1,2,…,23}Formula (2)

[0098] Among them, ts represents the timestamp corresponding to the current time step, and HoD(·) represents the function of calculating the hour code for ts. In the function expression, ts.hour represents the hour corresponding to the timestamp, and finally minus one converts the code into a base 0 index.

[0099] The weekly time series feature coding sequence is used to help the air traffic prediction model learn different change patterns in a week (such as significant traffic differences between weekdays and weekends), and its calculation formula is:

[0100]

[0101] Among them, ts represents the timestamp corresponding to the current time step, and DoW(·) represents the function for calculating the week code for ts. In the function expression, ts.day represents the date in the month corresponding to the timestamp. Similarly, the code is converted to a base 0 index by subtracting 1 at the end. It should be understood that the day of the week here has no fixed correspondence with the traditional day of the week. It is just an index representation used to identify the degree of deviation based on the 7-day scale.

[0102] As a refinement of the above embodiment, when executing step 103 to decompose the normalized historical air traffic based on Fourier transform to obtain a time-invariant sequence and a time-varying sequence, the following implementation methods may also be used but are not limited to, for example: transforming the normalized time domain sequence of the historical air traffic into a frequency domain complex sequence based on discrete Fourier transform; filtering the high frequency components of the frequency domain complex sequence based on a low-pass filter, and converting the filtering result from the frequency domain to the time domain to obtain the time-invariant sequence; calculating the difference between the time domain sequence and the time-invariant sequence to obtain the time-varying sequence.

[0103] To facilitate understanding of the above embodiment, the discrete Fourier transform process can be described in combination with the formula:

[0104] Y=DFT(X) formula (4)

[0105] Wherein, DFT(·) represents discrete Fourier transform, X is the time domain sequence, and Y is the frequency domain complex sequence; further, formula (4) can be specifically expanded to formula (5) as follows:

[0106]

[0107] Among them, for X={x1,x2,…,x L} Perform discrete Fourier transform and get L is the length of the time domain sequence, ω k is the normalized frequency parameter, i is the imaginary unit, k is the index of the frequency domain sequence, that is, the frequency domain complex sequence, e is the natural logarithm, and n is the index of traversal X.

[0108] The high-frequency components of the frequency domain complex sequence are filtered based on a low-pass filter, and the filtering result is converted from the frequency domain to the time domain. For the process of obtaining the time-invariant sequence, please refer to formula (6), formula (7), and formula (8):

[0109] Filter out the high-frequency components of the frequency domain complex sequence Y, and then convert it back to the time domain to obtain the time-invariant sequence X inv :

[0110] X inv=iDFT(Y⊙LPF) formula (6)

[0111] Where LPF is a pre-built low-pass filter, ⊙ represents the Hadamard product, i.e., element-by-element multiplication, and iDFT(·) represents the inverse discrete Fourier transform.

[0112] The expression for the pre-built low-pass filter LPF is as follows:

[0113]

[0114] Among them, f c is the cut-off frequency, when the frequency component f k When it is greater than the threshold, the filter value is 0 and the frequency component is blocked; otherwise, the filter value is 1 and the frequency component is retained. k is the subscript index.

[0115] The formula for the inverse discrete Fourier transform is as follows:

[0116]

[0117] in, is the frequency component after low-pass filtering, is a time-invariant sequence converted to the time domain, L is the length of the time domain sequence, ω k is the normalized frequency parameter, i is the imaginary unit, k is the index of the frequency domain sequence, that is, the frequency domain complex sequence, e is the natural logarithm, and n is the index of traversing X. Please refer to the parameter interpretation of formula (3).

[0118] Time domain series X and time invariant series X inv The difference between the two is the time-varying sequence X var :

[0119] f var =XX inv Formula (9)

[0120] As a refinement of the above embodiment, when performing feature extraction on the time-invariant sequence and the time-varying sequence respectively to obtain the time-invariant features and the time-varying features as described in step 104, the following implementation methods may also be adopted but are not limited to, for example: linear mapping processing is performed on the time-invariant sequence and the time-varying sequence respectively to obtain an aligned time-invariant sequence aligned to a first preset sequence length and an aligned time-varying sequence aligned to a second preset sequence length; high-dimensional feature extraction is performed on the aligned time-invariant sequence and the aligned time-varying sequence respectively based on the convolution layer to obtain the time-invariant features and the time-varying features.

[0121] In order to facilitate the understanding of the above embodiment, this embodiment combines the above formulas (3)-(9) and explains the above embodiment in a formula form. The specific process is as follows:

[0122] The linear mapping is used to align the time-varying sequence and the time-invariant sequence to a predicted time sequence length, respectively, and the time sequence length includes the first preset sequence length and the second preset sequence length:

[0123] X′ var =Linear1(X var )Formula (10)

[0124] X′ inv =Linear2(X inv )Formula (11)

[0125] Among them, Linear1(·) and Linear2(·) are two independent linear layers. is an aligned time-varying sequence, is an aligned time-invariant sequence, L represents the length of the historical sequence, T represents the length of the predicted sequence, Represents a parameter data set of (L+T)×1 dimension;

[0126] The convolutional layer embeds the aligned time-varying sequence and the aligned time-invariant sequence into a high-dimensional space, i.e., a representation of high-dimensional feature extraction:

[0127] X″ var =Conv1D(X′ var )Formula (12)

[0128] X″ inv =Conv1S(X′ inv )Formula (13)

[0129] Among them, Conv1D(·) is a one-dimensional convolutional layer with shared weights, is the time-varying sequence embedding, i.e., the time-varying feature, It is the time-invariant sequence embedding, that is, the time-invariant feature, L represents the length of the historical sequence, T represents the length of the predicted sequence, Represents (L+T)×D seq Dimensional parameter dataset, D seq The dimension of embedding time-varying sequences and embedding time-invariant sequences into high-dimensional space.

[0130] As a refinement of the above embodiment, when executing step 105 to perform data splicing processing on the multi-resolution time series feature coding sequence with the time-invariant features and the time-varying features respectively to obtain time series enhanced time-varying sequence embedded data and time series enhanced time-invariant sequence embedded data, the following implementation methods may also be adopted but are not limited to, for example: high-dimensional feature extraction is performed on the moment time series feature coding sequence, the hour time series feature coding sequence, and the week time series feature coding sequence, and the three feature extraction results are spliced ​​to obtain a potential time series representation; the potential time series representation is spliced ​​with the time-invariant features to obtain the time series enhanced time-invariant sequence embedded data; the potential time series representation is spliced ​​with the time-varying features to obtain the time series enhanced time-varying sequence embedded data.

[0131] In order to facilitate understanding of the above embodiment, this embodiment combines the above formulas (3)-(13) and explains the above embodiment in a formula form. The specific process is as follows:

[0132] The three multi-resolution time series feature coding sequences generated in step 102, namely, the moment time series feature coding sequence Q, the hour time series feature coding sequence H, and the week time series feature coding sequence D, are embedded into a high-dimensional space and concatenated to obtain a potential time series representation LTR:

[0133] LTR=Concat(Emb1(Q),Emb2(H),Emb3(D)) formula (14)

[0134] Among them, Emb1(·), Emb2(·), and Emb3(·) represent three independent embedding layers, and their dimensions are D q ,D h ,D d , Concat(·) represents the concatenation operation along the feature dimension,

[0135] Then, embed the time-varying sequence into X″ var and time-invariant temporal embedding X″ inv To better preserve the time series information, we use the splicing fusion method along the feature dimension to obtain the time-varying sequence embedding with enhanced time series features, also known as the time-varying sequence embedding data E. var and the time-series feature enhanced time-invariant sequence embedding, that is, the time-series enhanced time-invariant sequence embedding data E inv :

[0136] E var =Concat(X″) var ,LTR) Formula (15)

[0137] E inv =Concat(X″)inv ,LTR) Formula (16)

[0138] Among them, Concat(·) represents the concatenation operation along the feature dimension.

[0139] As a refinement of the above embodiment, when performing step 106 of learning the time-varying sequence embedding data and the time-invariant sequence embedding data based on the long short-term memory network to obtain the time-invariant time series features and the time-varying time series features, the following implementation methods may be adopted but are not limited to, for example: the long short-term memory network is used to learn the time series association between the air traffic flows to obtain the high-dimensional time-varying representation after the time series feature extraction, that is, the time-varying time series feature H var and high-dimensional time-invariant representation, namely the time-invariant time series feature H inv :

[0140] H var =LSTM(E var )Formula (17)

[0141] H inv =LSTM(E inv )Formula (18)

[0142] Among them, LSTM(·) is a long short-term memory network with hidden layer dimension D.

[0143] As a refinement of the above embodiment, when performing weighted processing based on the attention scores corresponding to the time-invariant temporal features and the time-varying temporal features in step 107 to obtain a weighted time-invariant temporal feature sequence and a weighted time-varying temporal feature, the following implementation methods may also be used but are not limited to, for example: capturing the temporal dependency association of the high-dimensional representation based on the attention mechanism to obtain a high-dimensional semantic time-varying representation weighted by the attention score, that is, the weighted time-varying temporal feature sequence A var and the high-dimensional semantic time-invariant representation, i.e., the weighted time-invariant temporal feature sequence A inv :

[0144] A var =AttBlock(H var )Formula (19)

[0145] A inv =AttBlock(H inv )Formula (20)

[0146] Among them, AttBlock(·) represents the attention module,

[0147] Specifically, for the attention module Attblock, the calculation process of formula (19) and formula (20) is as follows:

[0148] Calculate the score S of the feature map for the high-dimensional representation H f , H is H inv and H var Any of:

[0149] S f =γ(H·W f1 )·W f2 Formula (21)

[0150] in, is the weight matrix for feature mapping, and γ(·) represents the ReLU activation function;

[0151] Weigh the importance of each time step and get the context-aware attention score S t :

[0152] S t =σ(W t ·S f )Formula (22)

[0153] in, is the weight matrix used to enhance temporal dependency, σ(·) represents the Sigmoid activation function;

[0154] Weighted by the attention score, we get a high-dimensional semantic representation A:

[0155]

[0156] in, Indicates dimensional expansion according to the broadcast mechanism, H is H inv and H var Any one of them, A is the A corresponding to H var or A inv .

[0157] As a refinement of the above embodiment, in executing step 108 and step 109, the KAN prediction network is used to perform air traffic forecasting on the weighted time-invariant time series feature sequence and the weighted time-varying time series feature to obtain a predicted time-invariant air traffic sequence and a predicted time-varying air traffic sequence; the sum of the predicted time-invariant air traffic sequence and the predicted time-varying air traffic sequence is denormalized to obtain the actual air traffic forecast result, including: mapping the high-dimensional semantic representation to a one-dimensional sequence based on the KAN prediction network to obtain a predicted time-varying sequence, i.e., the predicted time-varying air traffic sequence P var and the time-invariant sequence, namely the predicted time-invariant aviation traffic sequence P inv :

[0158] P var =KANLayer(A var )Formula (24)

[0159] P inv =KANLayer(A inv )Formula (25)

[0160] Wherein, KANLayer(·) represents the Kolmogorov-Arnold layer; different from the existing MLPLayer, the edge of the KANLayer has a learnable, nonlinear function φ(x), expressed as follows:

[0161] φ(x)=Spline(x)+SiLU(x) Formula (26)

[0162] Among them, Spline(·) represents the B-spline spline function, SiLU(·) represents the SiLU activation function as the basis, and x represents the value on the node corresponding to the edge.

[0163] Finally, the predicted time-varying sequence P var and time-invariant sequence P inv Add them together to get the final result P of traffic flow prediction:

[0164] P=P var +P inv Formula (27)

[0165] Wherein, P is subjected to a denormalization process to obtain the actual aviation traffic prediction result.

[0166] In order to facilitate understanding of the above embodiment, this embodiment provides a normalization and denormalization process. The principles of the normalization and denormalization involved in this disclosure are the same as those of this embodiment, including: the normalization method is z-score normalization, and the formula is:

[0167]

[0168] Among them, x represents the original value, μ represents the mean of the data set, σ represents the standard deviation of the data set, and z represents the normalized value.

[0169] The formula for denormalization is:

[0170] x=z·σ+μFormula (29)

[0171] Among them, x represents the original value, μ represents the mean of the data set, σ represents the standard deviation of the data set, and z represents the normalized value.

[0172] As a refinement of the above embodiment, when executing step 101 to receive the air traffic data and perform preprocessing to obtain the historical air traffic, the following implementation methods may also be used but are not limited to, for example: converting the format of the air traffic data obtained and updated in real time into a structured data format that can be used for subsequent processing; extracting key data fields from the air traffic data after format conversion, and removing duplicate data therein; constructing a time series based on the historical time step length based on the deduplicated key data fields to generate the historical air traffic. To facilitate understanding of this embodiment, please refer to the following exemplary descriptions in combination with actual scenarios, including: receiving track data through an air traffic management system, including but not limited to integrated track, primary / secondary radar, ADS-B and other data; pre-processing the received data (such as decoding the ADS-B raw data) and converting it into a structured data format for subsequent processing; extracting key data fields from the decoded data, including the flight number, location information (longitude, latitude, flight altitude), speed, heading, etc. of the aircraft; counting the number of flights in the airspace according to a predetermined time step and in combination with the real-time position of the aircraft, de-duplicating according to the unique identifier of the flight (such as the flight number), and eliminating invalid or incomplete data; defining a lookback window length L and a prediction time step T, constructing a time series according to the number of flights in the airspace at each time step, and generating historical air traffic data X 1:L ={x1,x2,…x L}; Format and standardize the historical traffic data to ensure that various traffic indicators meet the model input requirements; Store the historical traffic data for use by subsequent prediction models; wherein the track data is dynamically updated in real time and the frequency of recording the track data includes but is not limited to every 15 minutes.

[0173] As a refinement of the above embodiment, before executing step 101, the method can also adopt but is not limited to the following implementation methods, for example: training a preset deep learning framework to obtain an air traffic prediction model, and the air traffic prediction model is used to execute the air traffic prediction method; the air traffic prediction model includes a frequency perception module, a sequence modeling module and a KAN prediction module; the frequency perception module is used to decompose and extract features of the historical air traffic input into the air traffic prediction model to generate time series features, and the time series features include the time-invariant features and the time-varying features; the sequence modeling module used for enhancing the time series features is used to fuse the time series features with the multi-resolution time series feature encoding sequence, and extract the time series dependency of the fusion result based on the long short-term memory network; the KAN prediction module based on the attention mechanism is used to map the weighted time-invariant time series feature sequence and the weighted time-varying time series feature to prediction dimensions respectively and output a predicted air traffic sequence, and the predicted air traffic sequence includes the predicted time-invariant air traffic sequence and the predicted time-varying air traffic sequence.

[0174] In order to more clearly demonstrate the architecture of an air traffic prediction model provided by the above embodiment, please refer to Figures 2 to 4 , Figure 2 The overall structure diagram of an air traffic prediction model provided by an embodiment of the present disclosure is as follows: Figure 3 A structural diagram of a decomposition module provided in an embodiment of the present disclosure is shown in FIG. Figure 4 A schematic diagram of the structure of a KAN prediction module provided in an embodiment of the present disclosure. Figure 2 In the embodiment, the air traffic flow sequence is the historical air traffic flow, the air traffic flow prediction model includes the time series feature encoding module for executing step 102, the frequency perception module includes the decomposition module, the linear mapping and the convolution layer, the frequency perception module is used to execute the steps involved in formula (4) to formula (13), the sequence modeling module includes an embedding layer, a feature fusion module, and an LSTM network, the embedding layer is used for data splicing of the three multi-resolution time series feature encoding sequences, the feature fusion module is used to execute step 105, the LSTM network is used to execute step 106, the KAN prediction module based on the attention mechanism includes an attention module and a KAN prediction module, and the KAN prediction module based on the attention mechanism is used to execute the processes involved in formula (19) to formula (27). Figure 3 The decomposition module shown, whose internal structure corresponds to the process of data processing involved in formula (4) to formula (8), Figure 4 The KAN prediction module with attention mechanism shown in the figure has an internal structure for data processing, which can refer to the processing process of formula (19) to formula (27).

[0175] In some embodiments, the historical time step (L) length of the air traffic prediction model is set to 96. At the same time, the prediction time step (T) is set to five different lengths: {12, 24, 48, 96, 192}. That is, the observation data of the past day is used to predict the traffic flow of {3 hours, 6 hours, 12 hours, 1 day, 2 days}. In the frequency perception module, the cutoff frequency f c Set to 10% f s The convolutional layer Conv1D has 128 channels; the embedding dimensions of Emb1(·), Emb2(·), and Emb3(·) are 128, 64, and 64, respectively. The LSTM has 2 layers and 128 hidden units. The prediction network consists of two KANLayers, and the number of hidden units is [384, 192, 1]. The scaling factor α is set to 0.6.

[0176] As a refinement of the above embodiment, when the preset deep learning framework is trained to obtain the air traffic prediction model in the execution step, the following implementation methods may also be adopted but are not limited to, for example: the air traffic prediction model is trained using the constructed training set, the network parameters of the air traffic prediction model are updated by back propagation using the loss function, and the learning rate is dynamically adjusted using the Adam optimizer, and the formula of the loss function is expressed as follows:

[0177] L dd =L time +αL freq Formula (30)

[0178] Among them, L dd is the loss function, L time is the time domain loss, L freq is the frequency domain loss, α represents the scaling factor, which is used to control the impact of the frequency domain loss on the total loss;

[0179] The calculation formula of the time domain loss is as follows:

[0180]

[0181] Among them, x ij represents the true value of the time domain series, represents the predicted value of the time domain sequence, T represents the length of the predicted time domain sequence, N is the number of time domain sequence samples in a training batch, i and j represent subscript variables, i is the length index, and j is the sample index;

[0182] The frequency domain loss is calculated as follows:

[0183] y=DFT(x),

[0184]

[0185] Where DFT(·) represents discrete Fourier transform, x is the time domain sequence, y is the frequency domain sequence, To predict the time domain series, To predict the frequency domain sequence, y ij is the true value of the frequency domain sequence, is the predicted value of the frequency domain series, represents the length of the frequency domain sequence after real Fourier transform, T represents the length of the predicted time domain sequence, n is the number of frequency domain sequence samples in a training batch, i and j represent subscript variables, i is the length index, and j is the sample index.

[0186] In summary, the embodiments of the present disclosure can achieve the following effects:

[0187] 1. The multi-resolution time series feature coding sequence is generated by performing multi-resolution time series feature coding on the historical air traffic, and the multi-resolution time series feature coding sequence includes significant multi-resolution time series features corresponding to the historical air traffic. At the same time, the multi-resolution time series feature coding sequence is respectively subjected to data splicing processing with the time-invariant feature and the time-varying feature to achieve time series feature enhancement of the time-varying feature and the time-invariant feature, so that the air traffic prediction model can more easily capture the time series correlation of air traffic changes. In addition, the normalized historical air traffic is decomposed based on Fourier transform to obtain time-invariant sequence and time-varying sequence, so as to realize the perception of the air traffic prediction model to different time series frequencies. Furthermore, by integrating the aforementioned mechanism into the air traffic prediction model, the performance breakthrough of the air traffic prediction model and the improvement of the accuracy of the prediction results are achieved.

[0188] Corresponding to the above-mentioned method for predicting air traffic, the present invention further proposes an apparatus for predicting air traffic. Since the apparatus embodiment of the present invention corresponds to the above-mentioned method embodiment, details not disclosed in the apparatus embodiment can be referred to the above-mentioned method embodiment, and will not be described in detail in the present invention.

[0189] Figure 5 A schematic diagram of the structure of an aviation traffic prediction device provided by an embodiment of the present disclosure is shown in FIG. Figure 5 As shown, including:

[0190] The receiving unit 21 is used to receive and pre-process the air traffic data to obtain the historical air traffic;

[0191] The encoding unit 22 is used to perform multi-resolution time series feature encoding on the historical air traffic to generate a multi-resolution time series feature encoding sequence;

[0192] A processing unit 23 is used to decompose the normalized historical air traffic based on Fourier transform to obtain a time-invariant sequence and a time-varying sequence;

[0193] An extraction unit 24, configured to extract features from the time-invariant sequence and the time-varying sequence respectively to obtain a time-invariant feature and a time-varying feature;

[0194] A splicing unit 25 is used to perform data splicing processing on the multi-resolution time series feature coding sequence with the time-invariant feature and the time-varying feature respectively to obtain time series enhanced time-varying sequence embedded data and time series enhanced time-invariant sequence embedded data;

[0195] A learning unit 26 is used to learn the time-varying sequence embedding data and the time-invariant sequence embedding data based on a long short-term memory network to obtain a time-invariant time series feature and a time-varying time series feature;

[0196] A weighting unit 27, configured to perform weighted processing based on the attention scores corresponding to the time-invariant temporal features and the time-varying temporal features, respectively, to obtain a weighted time-invariant temporal feature sequence and a weighted time-varying temporal feature;

[0197] A prediction unit 28 is used to perform air traffic prediction on the weighted time-invariant time series feature sequence and the weighted time-varying time series feature based on a KAN prediction network, to obtain a predicted time-invariant air traffic sequence and a predicted time-varying air traffic sequence;

[0198] The denormalization unit 29 is used to perform denormalization processing on the sum of the predicted time-invariant air traffic sequence and the predicted time-varying air traffic sequence to obtain the actual air traffic prediction result.

[0199] The air traffic device provided by the present invention receives air traffic data and performs preprocessing to obtain historical air traffic; performs multi-resolution time series feature encoding on the historical air traffic to generate a multi-resolution time series feature encoding sequence; decomposes the normalized historical air traffic based on Fourier transform to obtain a time-invariant sequence and a time-varying sequence; performs feature extraction on the time-invariant sequence and the time-varying sequence to obtain a time-invariant feature and a time-varying feature; performs data splicing processing on the multi-resolution time series feature encoding sequence with the time-invariant feature and the time-varying feature to obtain time-series enhanced time-varying sequence embedded data and time-series enhanced time-invariant sequence embedded data; based on long-term and short-term memory The memory network learns the time-varying sequence embedding data and the time-invariant sequence embedding data respectively to obtain the time-invariant time series features and the time-varying time series features; based on the attention scores corresponding to the time-invariant time series features and the time-varying time series features, weighted processing is performed respectively to obtain the weighted time-invariant time series feature sequence and the weighted time-varying time series feature; based on the KAN prediction network, the weighted time-invariant time series feature sequence and the weighted time-varying time series feature are used to predict air traffic flow respectively to obtain the predicted time-invariant air traffic flow sequence and the predicted time-varying air traffic flow sequence; the sum of the predicted time-invariant air traffic flow sequence and the predicted time-varying air traffic flow sequence is denormalized to obtain the actual air traffic prediction result. Compared with the related art, the multi-resolution time series feature coding sequence is generated by performing multi-resolution time series feature coding on the historical air traffic, and the multi-resolution time series feature coding sequence includes significant multi-resolution time series features corresponding to the historical air traffic. At the same time, the multi-resolution time series feature coding sequence is respectively subjected to data splicing processing with the time-invariant feature and the time-varying feature to achieve time series feature enhancement of the time-varying feature and the time-invariant feature, so that the air traffic prediction model can more easily capture the time series correlation of air traffic changes. In addition, the normalized historical air traffic is decomposed based on Fourier transform to obtain time-invariant sequence and time-varying sequence, so as to realize the perception of the air traffic prediction model to different time series frequencies. Furthermore, by integrating the aforementioned mechanism into the air traffic prediction model, the performance breakthrough of the air traffic prediction model and the improvement of the accuracy of the prediction results are achieved.

[0200] It should be noted that the above explanation of the method embodiment is also applicable to the device of this embodiment, and the principle is the same, which is not limited in this embodiment.

[0201] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0202] Figure 6A schematic block diagram of an example electronic device 300 that can be used to implement an embodiment of the present disclosure 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 electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0203] like Figure 6 As shown, the device 300 includes a computing unit 301, which can perform various appropriate actions and processes according to a computer program stored in a ROM (Read-Only Memory) 302 or a computer program loaded from a storage unit 308 to a RAM (Random Access Memory) 303. In the RAM 303, various programs and data required for the operation of the device 300 can also be stored. The computing unit 301, the ROM 302, and the RAM 303 are connected to each other via a bus 304. An I / O (Input / Output) interface 305 is also connected to the bus 304.

[0204] A number of components in the device 300 are connected to the I / O interface 305, including: an input unit 306, such as a keyboard, a mouse, etc.; an output unit 307, such as various types of displays, speakers, etc.; a storage unit 308, such as a disk, an optical disk, etc.; and a communication unit 309, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 309 allows the device 300 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0205] The computing unit 301 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 301 include, but are not limited to, a CPU (Central Processing Unit), a GPU (Graphic Processing Units), various dedicated AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, a DSP (Digital Signal Processor), and any appropriate processor, controller, microcontroller, etc. The computing unit 301 performs the various methods and processes described above, such as the prediction method of air traffic. For example, in some embodiments, the prediction method of air traffic may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 308. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 300 via the ROM 302 and / or the communication unit 309. When the computer program is loaded into the RAM 303 and executed by the computing unit 301, one or more steps of the method described above may be performed. Alternatively, in other embodiments, the computing unit 301 may be configured to execute the aforementioned aviation traffic prediction method in any other appropriate manner (for example, by means of firmware).

[0206] Various embodiments of the systems and techniques described above herein may be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application Specific Standard Products), SOCs (System On Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include: being implemented in one or more computer programs that may be executed and / or interpreted on a programmable system including at least one programmable processor that may be a dedicated or general-purpose programmable processor that may 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.

[0207] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0208] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage media may include electrical connections based on one or more lines, portable computer disks, hard disks, RAM, ROM, EPROM (Electrically Programmable Read-Only-Memory) or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0209] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer 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 computer. 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).

[0210] The systems and techniques described herein may be implemented in a computing system that includes backend 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 frontend components (e.g., a user computer with a graphical user interface or a 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 backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: LAN (Local Area Network), WAN (Wide Area Network), the Internet, and blockchain networks.

[0211] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0212] It should be noted that artificial intelligence is a discipline that studies how computers can simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.), and includes both hardware-level and software-level technologies. Artificial intelligence hardware technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, and big data processing; artificial intelligence software technologies mainly include computer vision technology, speech recognition technology, natural language processing technology, as well as machine learning / deep learning, big data processing technology, knowledge graph technology, and other major directions.

Claims

1. A method for predicting air traffic, characterized in that: include: Receive air traffic data and pre-process it to obtain historical air traffic; Performing multi-resolution time series feature encoding on the historical air traffic to generate a multi-resolution time series feature encoding sequence; Decomposing the normalized historical air traffic based on Fourier transform to obtain a time-invariant sequence and a time-varying sequence; Extracting features from the time-invariant sequence and the time-varying sequence respectively to obtain time-invariant features and time-varying features; The multi-resolution time series feature coding sequence is respectively subjected to data splicing processing with the time-invariant feature and the time-varying feature to obtain time series enhanced time-varying sequence embedded data and time series enhanced time-invariant sequence embedded data; Based on the long short-term memory network, the time-varying sequence embedding data and the time-invariant sequence embedding data are learned respectively to obtain the time-invariant time series features and the time-varying time series features; Based on the attention scores corresponding to the time-invariant temporal features and the time-varying temporal features, weighted processing is performed respectively to obtain a weighted time-invariant temporal feature sequence and a weighted time-varying temporal feature; Based on the KAN prediction network, the air traffic flow is predicted for the weighted time-invariant time series feature sequence and the weighted time-varying time series feature, and the predicted time-invariant air traffic flow sequence and the predicted time-varying air traffic flow sequence are obtained. The sum of the predicted time-invariant aviation traffic sequence and the predicted time-varying aviation traffic sequence is subjected to denormalization processing to obtain an actual aviation traffic prediction result.

2. The method according to claim 1, characterized in that The performing multi-resolution time series feature encoding on the historical air traffic to generate a multi-resolution time series feature encoding sequence comprises: Performing time series feature encoding with time resolution on the historical air traffic to obtain a time series feature encoding sequence; Performing time series feature encoding with hourly resolution on the historical air traffic to obtain an hourly time series feature encoding sequence; The historical air traffic is encoded with time series features at a weekly resolution to obtain a weekly time series feature coding sequence, wherein the multi-resolution time series feature coding sequence includes the moment time series feature coding sequence, the hour time series feature coding sequence and the weekly time series feature coding sequence.

3. The method according to claim 1, characterized in that: The decomposition process of the normalized historical air traffic based on Fourier transform to obtain the time-invariant sequence and the time-varying sequence includes: Transforming the normalized time domain sequence of the historical air traffic into a frequency domain complex sequence based on discrete Fourier transform; Filtering the high-frequency components of the frequency-domain complex sequence based on a low-pass filter, and converting the filtering result from the frequency domain to the time domain to obtain the time-invariant sequence; The difference between the time domain sequence and the time invariant sequence is calculated to obtain the time varying sequence.

4. The method according to claim 2, characterized in that: The extracting features of the time-invariant sequence and the time-varying sequence respectively to obtain the time-invariant features and the time-varying features comprises: Performing linear mapping processing on the time-invariant sequence and the time-varying sequence respectively to obtain an aligned time-invariant sequence aligned to a first preset sequence length and an aligned time-varying sequence aligned to a second preset sequence length; Based on the convolution layer, high-dimensional features are extracted for the aligned time-invariant sequence and the aligned time-varying sequence respectively to obtain the time-invariant features and the time-varying features.

5. The method according to claim 4, characterized in that The step of performing data splicing processing on the multi-resolution time series feature coding sequence with the time-invariant feature and the time-varying feature to obtain time series enhanced time-varying sequence embedded data and time series enhanced time-invariant sequence embedded data comprises: Performing high-dimensional feature extraction on the moment time series feature coding sequence, the hour time series feature coding sequence, and the week time series feature coding sequence, and concatenating the three feature extraction results to obtain a potential time series representation; Concatenating the latent temporal representation with the time-invariant feature to obtain the temporal enhanced time-invariant sequence embedding data; The latent temporal representation and the time-varying feature are concatenated to obtain the temporal enhanced time-varying sequence embedding data.

6. The method according to claim 1, characterized in that The receiving of air traffic data and preprocessing to obtain historical air traffic includes: Converting the format of the air traffic data acquired and updated in real time into a structured data format suitable for subsequent processing; Extract key data fields from the air traffic data after format conversion and remove duplicate data; A time series is constructed based on the deduplicated key data fields according to the historical time step length to generate the historical aviation traffic.

7. The method according to any one of claims 1 to 6, characterized in that: Before receiving and preprocessing the air traffic data to obtain historical air traffic, the method includes: Training a preset deep learning framework to obtain an air traffic prediction model, wherein the air traffic prediction model is used to execute the air traffic prediction method; The aviation traffic prediction model includes a frequency perception module, a sequence modeling module and a KAN prediction module; The frequency perception module is used to decompose and extract features of the historical air traffic input into the air traffic prediction model to generate time series features, wherein the time series features include the time-invariant features and the time-varying features; The sequence modeling module for enhancing the temporal features is used to fuse the temporal features with the multi-resolution temporal feature encoding sequence, and extract temporal dependency from the fusion result based on the long short-term memory network; The KAN prediction module based on the attention mechanism is used to map the weighted time-invariant temporal feature sequence and the weighted time-varying temporal feature to prediction dimensions respectively and output a predicted air traffic sequence, wherein the predicted air traffic sequence includes the predicted time-invariant air traffic sequence and the predicted time-varying air traffic sequence.

8. The method according to claim 7, characterized in that The training of the preset deep learning framework to obtain the aviation traffic prediction model includes: The constructed training set is used to train the air traffic prediction model, and the loss function is used for back propagation to update the network parameters of the air traffic prediction model. The Adam optimizer is used to dynamically adjust the learning rate. The formula of the loss function is expressed as follows: L dd =L time +αL freq Among them, L dd is the loss function, L time is the time domain loss, L freq is the frequency domain loss, α represents the scaling factor, which is used to control the impact of the frequency domain loss on the total loss; The calculation formula of the time domain loss is as follows: Among them, x ij represents the true value of the time domain series, represents the predicted value of the time domain sequence, T represents the length of the predicted time domain sequence, N is the number of time domain sequence samples in a training batch, i and j represent subscript variables, i is the length index, and j is the sample index; The frequency domain loss is calculated as follows: y=DFT(x), Where DFT(·) represents discrete Fourier transform, x is the time domain sequence, y is the frequency domain sequence, To predict the time domain series, To predict the frequency domain sequence, y ij is the true value of the frequency domain sequence, is the predicted value of the frequency domain series, represents the length of the frequency domain sequence after the real Fourier transform, T represents the length of the predicted time domain sequence before the real Fourier transform, n is the number of frequency domain sequence samples in a training batch, i and j represent subscript variables, i is the length index, and j is the sample index.

9. An air traffic prediction device, characterized in that: include: A receiving unit, used for receiving and preprocessing the air traffic data to obtain the historical air traffic; An encoding unit, used for performing multi-resolution time series feature encoding on the historical air traffic to generate a multi-resolution time series feature encoding sequence; A processing unit, configured to decompose the normalized historical air traffic based on Fourier transform to obtain a time-invariant sequence and a time-varying sequence; An extraction unit, used to extract features from the time-invariant sequence and the time-varying sequence respectively to obtain a time-invariant feature and a time-varying feature; A splicing unit, used for performing data splicing processing on the multi-resolution time series feature coding sequence with the time-invariant feature and the time-varying feature respectively, to obtain time series enhanced time-varying sequence embedded data and time series enhanced time-invariant sequence embedded data; A learning unit, used for learning the time-varying sequence embedding data and the time-invariant sequence embedding data based on a long short-term memory network, respectively, to obtain a time-invariant time series feature and a time-varying time series feature; A weighting unit, configured to perform weighted processing based on the attention scores corresponding to the time-invariant temporal features and the time-varying temporal features, respectively, to obtain a weighted time-invariant temporal feature sequence and a weighted time-varying temporal feature; A prediction unit is used to perform air traffic prediction on the weighted time-invariant time series feature sequence and the weighted time-varying time series feature based on a KAN prediction network, to obtain a predicted time-invariant air traffic sequence and a predicted time-varying air traffic sequence; The denormalization unit is used to perform denormalization processing on the sum of the predicted time-invariant aviation traffic sequence and the predicted time-varying aviation traffic sequence to obtain the actual aviation traffic prediction result.

10. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 6.

Images