Traffic prediction method, device, equipment and medium based on big data

Abstract

The present invention relates to the field of data processing technology, and discloses a traffic prediction method, device, equipment and medium based on big data, including: obtaining an unlabeled sample data set, a labeled sample data set and a current business data set; performing feature replacement processing on each unlabeled sample data in the unlabeled sample data set to obtain a replacement feature sample data set; inputting each replacement feature sample data in the replacement feature sample data set into an initial traffic prediction model for model pre-training to obtain a reference traffic prediction model; inputting each labeled sample data in the labeled sample data set into a reference traffic prediction model for model parameter adjustment processing to obtain a target traffic prediction model; inputting the current business data set into the target traffic prediction model for traffic prediction to obtain a target traffic prediction result, which can obtain a more accurate traffic prediction result.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a traffic prediction method, device, equipment and medium based on big data. Background Art

[0002] With the construction and widespread adoption of 5G networks, user demand for high-speed data transmission continues to increase. The rapid growth of 5G network traffic has made traffic prediction for 5G networks a hot research topic. Currently, various machine learning and data mining techniques are commonly used to analyze and model data traffic in 5G networks to predict future traffic flows. However, as data volumes increase, traditional models are increasingly unable to absorb knowledge, resulting in inaccurate traffic prediction results for 5G networks. Therefore, achieving more accurate traffic prediction results has become a pressing issue. Summary of the Invention

[0003] The present invention provides a traffic prediction method, device, equipment and medium based on big data to solve the technical problem of how to obtain more accurate traffic prediction results.

[0004] In a first aspect, a traffic prediction method based on big data is provided, comprising:

[0005] Obtaining an unlabeled sample data set, a labeled sample data set, and a current business data set;

[0006] Perform feature replacement processing on each unlabeled sample data in the unlabeled sample data set to obtain a replacement feature sample data set;

[0007] Input each replacement feature sample data in the replacement feature sample data set into the initial traffic prediction model for model pre-training to obtain a reference traffic prediction model;

[0008] Input each labeled sample data in the labeled sample data set into the reference flow prediction model to adjust the model parameters to obtain the target flow prediction model;

[0009] The current business data set is input into the target traffic prediction model to perform traffic prediction and obtain the target traffic prediction result.

[0010] In a second aspect, a traffic prediction device based on big data is provided, comprising:

[0011] An acquisition module is used to acquire an unlabeled sample data set, a labeled sample data set, and a current business data set;

[0012] A first processing module is configured to perform feature replacement processing on each unlabeled sample data in the unlabeled sample data set to obtain a replacement feature sample data set;

[0013] The second processing module is used to input each replacement feature sample data in the replacement feature sample data set into the initial traffic prediction model for model pre-training to obtain a reference traffic prediction model;

[0014] A third processing module is used to input each labeled sample data in the labeled sample data set into the reference flow prediction model to adjust the model parameters and obtain a target flow prediction model;

[0015] The fourth processing module is used to input the current business data set into the target traffic prediction model to perform traffic prediction and obtain a target traffic prediction result.

[0016] In a third aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned traffic prediction method based on big data are implemented.

[0017] In a fourth aspect, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the steps of the above-mentioned big data-based traffic prediction method are implemented.

[0018] In the scheme implemented by the above-mentioned big data-based traffic prediction method, device, equipment and medium, by obtaining an unlabeled sample data set, a labeled sample data set and a current business data set, each unlabeled sample data in the unlabeled sample data set is subjected to feature replacement processing to obtain a replacement feature sample data set, so that each replacement feature sample data in the replacement feature sample data set can be input into the initial traffic prediction model for model pre-training to obtain a reference traffic prediction model, and each labeled sample data in the labeled sample data set is input into the reference traffic prediction model for model parameter adjustment processing to obtain a more accurate target traffic prediction model, and then the current business data set can be input into the target traffic prediction model for traffic prediction to obtain a more accurate target traffic prediction result, which is conducive to obtaining a more accurate traffic prediction result. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments of the present invention. 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 paying any creative labor.

[0020] Figure 1 This is a schematic diagram of an application environment of a traffic prediction method based on big data in one embodiment of the present invention;

[0021] Figure 2 This is a flow chart of a traffic prediction method based on big data in one embodiment of the present invention;

[0022] Figure 3 1 is a schematic structural diagram of a traffic prediction device based on big data in one embodiment of the present invention;

[0023] Figure 4 is a structural diagram of a computer device in one embodiment of the present invention;

[0024] Figure 5 FIG. 2 is another structural diagram of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0026] The terms "first," "second," and the like in the specification and claims of this application and the accompanying drawings are used to distinguish between different objects, not to describe a particular order. Furthermore, the terms "including," "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not limited to the listed steps or elements but may optionally include steps or elements not listed, or may optionally include other steps or elements inherent to the process, method, product, or apparatus.

[0027] References to "embodiments" in this application mean that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described in this application may be combined with other embodiments.

[0028] Current traffic forecasting technologies typically employ various machine learning and data mining techniques, such as time series analysis, regression analysis, and support vector machines, to combine multiple data sources, including historical traffic data, user behavior data, and network topology data, to predict and analyze future traffic trends. However, as data volumes increase, not only does the cost of labeling data continue to rise, but traditional models also struggle to utilize both labeled and unlabeled data. This means traditional models are unable to leverage existing data for deeper analysis, potentially leading to inaccurate traffic forecasts.

[0029] The traffic prediction method based on big data provided by the embodiment of the present invention can be applied in Figure 1 In an application environment, the client communicates with the server through a network. For example, taking the scenario of training the initial traffic prediction model based on the unlabeled sample data set and the labeled sample data set to obtain the target traffic prediction model, and performing traffic prediction based on the target traffic prediction model through the current business data set as an example, the target user can upload the unlabeled sample data set, the labeled sample data set and the current business data set through the client. The target user can be a manager who performs traffic prediction operations and / or a developer of the traffic prediction model, and this application does not impose any restrictions on this. Accordingly, the server can obtain the unlabeled sample data set, the labeled sample data set and the current business data set uploaded by the target user through the client, and perform feature replacement processing on each unlabeled sample data in the unlabeled sample data set to obtain a replacement feature sample data set, and input each replacement feature sample data in the replacement feature sample data set into the initial traffic prediction model for model pre-training to obtain a reference traffic prediction model, thereby inputting each labeled sample data in the labeled sample data set into the reference traffic prediction model for model parameter adjustment processing to obtain a target traffic prediction model, and then inputting the current business data set into the target traffic prediction model for traffic prediction to obtain a target traffic prediction result, and feeding back the target traffic prediction result to the client. Accordingly, the client can receive the target traffic prediction result from the server, and can display the target traffic prediction result on the client for the target user to query or browse. By adopting the traffic prediction method based on big data provided by the present application, the accuracy of the traffic prediction result can be improved.

[0030] The client can be, but is not limited to, various personal computers, laptops, smartphones, tablet computers, and portable wearable devices. The server can be implemented as an independent server or a server cluster consisting of multiple servers. The present invention is described in detail below through specific embodiments.

[0031] See also Figure 2 As shown, Figure 2A flow chart of a traffic prediction method based on big data provided by an embodiment of the present invention includes the following steps:

[0032] S10: Obtain an unlabeled sample data set, a labeled sample data set, and a current business data set.

[0033] The unlabeled sample data set may include one or more unlabeled sample data, which may be sample data in the historical business data set that is not labeled. The labeled sample data set may include one or more labeled sample data, which may be sample data in the historical business data set that is labeled. The historical business data set may include one or more historical business data, which may be business data generated in a historical time period. The current business data set may include one or more current business data, which may be business data generated in the current time period.

[0034] Optionally, the service data (such as current service data or historical service data) may include but is not limited to user number, user gender (0 represents male, 1 represents female), user age, user star rating (divided into five star ratings from 1 to 5), online time, average revenue brought by users in the current month (ARPU), ARPU of last month, ARPU of the month before yesterday, average monthly Internet traffic per household in the current month (dou), dou of last month, dou of the month before yesterday, average monthly talk time per household in the current month (mou), mou of last month, mou of the month before yesterday, average ARPU of the past three months, average dou of the past three months, average mou of the past three months, amount of voice call excess in the current month, amount of voice call excess in the previous month, amount of data traffic excess in the current month, amount of data traffic excess in the previous month, amount of data traffic excess in the month before yesterday, whether the user is a broadband user of this network (0 for no, 1 for yes), whether the user is a broadband user of another network : User (0 for no, 1 for yes), broadband bandwidth, whether broadband is activated (0 for no, 1 for yes), broadband bundling contract identifier (0 for not bundled, 1 for bundled), terminal bundling contract identifier (0 for not bundled, 1 for bundled), call fee contract identifier (0 for not contracted, 1 for contracted), package contract identifier (0 for not contracted, 1 for contracted), user's total package value, user's main tariff package (seven packages identified by 1 to 7), user traffic saturation this month, user traffic saturation last month, user traffic saturation the month before last, whether it is a home user, terminal type (three types identified by 1 to 3), whether the number preservation user is offset this month (0 for no, 1 for yes), whether the phone is changed this month (0 for no, 1 for yes), whether the place of residence covers 5g identification (0 for no, 1 for yes), whether the place of work covers 5g identification (0 for no, 1 for yes), etc. This application does not impose any restrictions on this.

[0035] It should be noted that the server can obtain the above-mentioned labeled sample data set by labeling a part of the historical business data in the historical business data set, or the server can directly obtain a part of the labeled historical business data, and use the other part of the unlabeled historical business data as the unlabeled sample data set, so as to further train the initial traffic prediction model based on the unlabeled sample data set and the labeled sample data set, so that the initial traffic prediction model can realize the prediction of future business data by fitting the historical business data. This application does not impose any restrictions on this.

[0036] S20: Perform feature replacement processing on each unlabeled sample data in the unlabeled sample data set to obtain a replaced feature sample data set.

[0037] The replacement feature sample data set may include one or more replacement feature sample data, and the replacement feature sample data may be used to indicate sample data obtained after feature replacement processing is performed on unlabeled sample data.

[0038] The server can perform feature replacement processing on some samples in the unlabeled sample data set. For example, the server can perform feature replacement processing on 10% or 15% of the unlabeled sample data in the unlabeled sample data set to obtain a replacement feature sample data set. This application does not impose any restrictions on this. Optionally, the server can replace the feature values of the unlabeled sample data with null values, or by swapping the features of the unlabeled sample data; or the server can replace the feature values of a portion (such as 10%) of the unlabeled sample data with null values, and swap the features of another portion (such as 5%) of the unlabeled sample data to achieve the above feature replacement processing. This application does not impose any restrictions on this.

[0039] After the server performs feature replacement on some of the unlabeled sample data in the unlabeled sample data set, the server can mark the unlabeled sample data on which the feature replacement operation has been performed. For example, the server can use a mask to mark the unlabeled sample data whose feature values are replaced with null values. The server can use the original feature to mark the unlabeled sample data after the feature replacement, so that the unlabeled sample data on which the feature replacement operation has been performed can have a replacement feature label. This application does not impose any restrictions on this. In other words, the replacement feature sample data set obtained by the feature replacement process can include the replacement feature label corresponding to each unlabeled sample data on which the feature replacement operation has been performed, such as a replacement feature label subset. This application does not impose any restrictions on this.

[0040] S30: Input each replacement feature sample data in the replacement feature sample data set into the initial traffic prediction model for model pre-training to obtain a reference traffic prediction model.

[0041] The initial traffic prediction model can be understood as an untrained traffic prediction model. This initial model can be constructed based on a feature selection processing module, a feature embedding layer, and a fully connected hierarchy to automatically select features from massive amounts of data. By learning and fitting known data, it can achieve accurate predictions for new data.

[0042] The reference traffic prediction model can be understood as the traffic prediction model obtained after training with a replacement feature sample data set. It is understood that the server preprocesses the unlabeled sample data in the unlabeled sample data set to obtain a replacement feature sample data set with replacement feature labels. This replacement feature sample data set can then be used to perform self-supervised pre-training on the initial traffic prediction model, thereby obtaining the reference traffic prediction model.

[0043] It should be understood that inputting each replacement feature sample data in the replacement feature sample data set into the initial traffic prediction model for model pre-training to obtain a reference traffic prediction model refers to the process of pre-training the initial traffic prediction model to obtain a reference traffic prediction model. Step S30, that is, inputting each replacement feature sample data in the replacement feature sample data set into the initial traffic prediction model for model pre-training to obtain a reference traffic prediction model, includes the following steps:

[0044] S31: using the initial traffic prediction model, obtaining a replacement feature label subset, a value replacement feature sample data subset, and a category replacement feature sample data subset from the replacement feature sample data set, wherein each replacement feature label in the replacement feature label subset is used to indicate the original feature corresponding to each replacement feature sample data;

[0045] S32: using the initial traffic prediction model, embedding each category replacement feature sample data in the category replacement feature sample data subset to obtain an embedded replacement feature sample data set;

[0046] S33: using the initial traffic prediction model, concatenating each embedded replacement feature sample data in the embedded replacement feature sample data set with each value replacement feature sample data in the value replacement feature sample data subset to obtain a concatenated feature sample data set;

[0047] S34: using the initial traffic prediction model, performing feature recovery processing on each spliced feature sample data in the spliced feature sample data set to obtain a restored feature sample data set;

[0048] S35: Pre-train the initial traffic prediction model in a direction of reducing the difference between each restored feature sample data in the restored feature sample data set and each replacement feature label in the replacement feature label subset to obtain a reference traffic prediction model.

[0049] The replacement feature label subset may include one or more replacement feature labels, which may be used to indicate the original features corresponding to each replacement feature sample data. The numerical value replacement feature sample data subset may include one or more numerical value replacement feature sample data, which may be used to indicate numerical sample data in the replacement feature sample data. The category replacement feature sample data subset may include one or more category replacement feature sample data, which may be used to indicate categorical sample data in the replacement feature sample data.

[0050] The embedded replacement feature sample data set may include one or more embedded replacement feature sample data, which may be used to indicate the sample data obtained after embedding the category replacement feature sample data. Optionally, the server may implement the embedding operation of each category replacement feature sample data through the embedding layer in the initial traffic prediction model, which is not limited in this application.

[0051] The splicing feature sample data set may include one or more splicing feature sample data, which may be used to indicate the sample data obtained after splicing the embedded replacement feature sample data and the numerical replacement feature sample data. Optionally, the server may perform a concat operation to implement the splicing operation of the embedded replacement feature sample data and the numerical replacement feature sample data, and this application does not impose any restrictions on this.

[0052] The restored feature sample data set may include one or more restored feature sample data, and the restored feature sample data may be used to indicate sample data obtained after feature restoration processing is performed on the spliced feature sample data.

[0053] It should be understood that using the initial traffic prediction model to perform feature recovery processing on each piece of spliced feature sample data in the spliced feature sample data set to obtain a restored feature sample data set refers to the process of restoring features of the sample data after feature replacement processing. Step S34, i.e., using the initial traffic prediction model to perform feature recovery processing on each piece of spliced feature sample data in the spliced feature sample data set to obtain a restored feature sample data set, includes the following steps:

[0054] S341: Using the initial traffic prediction model, perform first scale feature selection processing on each splicing feature sample data in the splicing feature sample data set to obtain a first feature sample data set;

[0055] S342: using the initial traffic prediction model, performing residual connection processing on each spliced feature sample data in the spliced feature sample data set and each first feature sample data in the first feature sample data set to obtain a first residual feature sample data set;

[0056] S343: using the initial traffic prediction model, performing a second scale feature selection process on each first residual feature sample data in the first residual feature sample data set to obtain a second feature sample data set;

[0057] S344: using the initial traffic prediction model, performing residual connection processing on each first residual feature sample data in the first residual feature sample data set and each second feature sample data in the second feature sample data set to obtain a second residual feature sample data set;

[0058] S345: Using the initial traffic prediction model, perform full connection processing on each second residual feature sample data in the second residual feature sample data set to obtain a restored feature sample data set.

[0059] Among them, the first feature sample data set may include one or more first feature sample data, and the first feature sample data can be used to indicate the sample data obtained after the spliced feature sample data is subjected to the first scale feature selection processing. The first scale feature selection processing can be a coarse-grained (or global granularity) feature selection processing, or a fine-grained (or local granularity) feature selection processing, and this application does not impose any restrictions on this. Optionally, the server can use the feature selection processing module (such as the first feature selection processing module) in the initial traffic prediction model to implement the operation of performing the first scale feature selection processing on each spliced feature sample data, and this application does not impose any restrictions on this.

[0060] It should be understood that using the initial traffic prediction model to perform first-scale feature selection processing on each piece of spliced feature sample data in the spliced feature sample data set to obtain the first feature sample data set refers to the process of performing first-scale feature selection. Step S341, i.e., using the initial traffic prediction model to perform first-scale feature selection processing on each piece of spliced feature sample data in the spliced feature sample data set to obtain the first feature sample data set, includes the following steps:

[0061] S3411: Using the initial traffic prediction model, extracting first-scale attention features from each spliced feature sample data in the spliced feature sample data set to obtain a first attention feature sample data set;

[0062] S3412: Using the initial traffic prediction model, obtaining attention feature sample data that meets preset conditions from the first attention feature sample data set to obtain a first target attention feature sample data subset;

[0063] S3413: Using the initial traffic prediction model, perform dimensionality-increasing mapping processing on each first target attention feature sample data in the first target attention feature sample data subset to obtain a first high-dimensional feature sample data set;

[0064] S3414: Using the initial traffic prediction model, activate each first high-dimensional feature sample data in the first high-dimensional feature sample data set to obtain a first activated feature sample data set;

[0065] S3415: Using the initial traffic prediction model, perform dimensionality reduction mapping processing on each first activated feature sample data in the first activated feature sample data set to obtain a first low-dimensional feature sample data set;

[0066] S3416: Using the initial traffic prediction model, perform residual connection processing on each spliced feature sample data in the spliced feature sample data set and each low-dimensional feature sample data in the first low-dimensional feature sample data set to obtain the first feature sample data set.

[0067] The first attention feature sample data set may include one or more first attention feature sample data, which may be used to indicate sample data obtained after performing first-scale attention feature extraction on the spliced feature sample data. Optionally, the server may use the attention matrix in the first feature selection processing module to perform first-scale attention feature extraction on the spliced feature sample data by adding attention weights, and this application does not impose any restrictions on this.

[0068] The first target attention feature sample data subset may include one or more first target attention feature sample data, and the first target attention feature sample data may be used to indicate the attention feature sample data that meets the preset conditions in the first attention feature sample data set. The preset conditions may be set by the target user or by the system default setting, which is not limited in this application. Optionally, the server may use the k (k≥1) features with the highest weights as preset conditions, which is not limited in this application. Optionally, after verification in the actual model training process, when k is one-third of the number of first attention feature sample data in the first attention feature sample data set, a better model training effect can be obtained, which is not limited in this application.

[0069] The first high-dimensional feature sample data set may include one or more first high-dimensional feature sample data, and the first high-dimensional feature sample data may be used to indicate the sample data obtained after the first target attention feature sample data is subjected to dimensionality-upgrading mapping processing. Optionally, the server may use a linear layer (Linear) in the first feature selection processing module, such as a dimensionality-upgrading linear layer, to implement the operation of performing dimensionality-upgrading mapping processing on the first target attention feature sample data, and this application does not impose any restrictions on this.

[0070] The first activation feature sample data set may include one or more first activation feature sample data, which may be used to indicate sample data obtained after activation processing of the first high-dimensional feature sample data. Optionally, the server may use an activation function, such as a ReLU activation function, to implement the operation of activating the first high-dimensional feature sample data, and this application does not impose any restrictions on this.

[0071] The first low-dimensional feature sample data set may include one or more first low-dimensional feature sample data, and the first low-dimensional feature sample data may be used to indicate sample data obtained after the first activation feature sample data is subjected to dimensionality reduction mapping processing. Optionally, the server may use a linear layer (Linear) in the first feature selection processing module, such as a dimensionality reduction linear layer, to implement the operation of performing dimensionality reduction mapping processing on the first activation feature sample data, and this application does not impose any restrictions on this.

[0072] After obtaining the first low-dimensional feature sample data set, the server can perform a residual connection between each first low-dimensional feature sample data in the first low-dimensional feature sample data set and each spliced feature sample data in the spliced feature sample data set, so that the final output first feature sample data can retain the characteristics of the input data (i.e., the spliced feature sample data), thereby achieving better model training effects, which is conducive to obtaining more accurate traffic prediction results in the subsequent traffic prediction process.

[0073] For steps S3411-S3416, the server adopts the initial traffic prediction model, performs first-scale attention feature extraction on each spliced feature sample data in the spliced feature sample data set, and obtains the first attention feature sample data set, so as to obtain attention feature sample data that meets the preset conditions from the first attention feature sample data set, obtain the first target attention feature sample data subset, and perform dimensionality up-mapping processing on each first target attention feature sample data in the first target attention feature sample data subset to obtain the first high-dimensional feature sample data set, thereby activating each first high-dimensional feature sample data in the first high-dimensional feature sample data set to obtain the first activated feature sample data set, and further performing dimensionality down-mapping processing on each first activated feature sample data in the first activated feature sample data set to obtain the first low-dimensional feature sample data set, and then performing residual connection processing on each spliced feature sample data in the spliced feature sample data set and each low-dimensional feature sample data in the first low-dimensional feature sample data set to obtain the first feature sample data set, which is conducive to training a more accurate reference traffic prediction model based on the first feature sample data set in subsequent steps.

[0074] The first residual feature sample data set may include one or more first residual feature sample data, and the first residual feature sample data may be used to indicate the sample data obtained after the spliced feature sample data and the first feature sample data are subjected to residual connection processing. Optionally, the server may implement the operation of performing residual connection processing on the spliced feature sample data and the first feature sample data through the feature selection processing module (i.e., the above-mentioned first feature selection processing module) in the initial traffic prediction model, and this application does not impose any restrictions on this. In other words, the feature selection processing module may perform the operation of performing residual connection processing on the input data and the processed data, so that the final output data can retain the characteristics of the original input data, thereby achieving a better pre-training effect.

[0075] The second feature sample data set may include one or more second feature sample data, which may be used to indicate sample data obtained after subjecting the first residual feature sample data to a second-scale feature selection process. The second scale may be different from the first scale. For example, if the first-scale feature selection is coarse-grained feature selection, the second-scale feature selection may be fine-grained feature selection. Alternatively, if both the first-scale feature selection and the second-scale feature selection are coarse-grained feature selection, the first-scale feature selection may be a first-type coarse-grained feature selection, and the second-scale feature selection may be a second-type coarse-grained feature selection. This application does not impose any restrictions on this.

[0076] Optionally, the server can use a feature selection processing module (such as a second feature selection processing module) in the initial traffic prediction model to implement the operation of performing second-scale feature selection processing on each first residual feature sample data. This application does not impose any restrictions on this. It is understandable that the scale (i.e., the first scale) at which the second feature selection processing module performs feature selection processing is different from the scale (i.e., the second scale) at which the first feature selection processing module performs feature selection processing. This application does not impose any restrictions on this.

[0077] The server performs a second scale feature selection process on each first residual feature sample data in the first residual feature sample data set to obtain the relevant steps of the second feature sample data set. Please refer to the detailed description of the aforementioned server performing a first scale feature selection process on each spliced feature sample data in the spliced feature sample data set to obtain the first feature sample data set, which will not be repeated in this application.

[0078] The second residual feature sample data set may include one or more second residual feature sample data, and the second residual feature sample data may be used to indicate the sample data obtained after the first residual feature sample data and the second feature sample data are subjected to residual connection processing. Optionally, the server may implement the operation of performing residual connection processing on the first residual feature sample data and the second feature sample data through the feature selection processing module (i.e., the above-mentioned second feature selection processing module) in the initial traffic prediction model, and this application does not impose any restrictions on this.

[0079] This application is explained by taking the example of a server including two feature selection processing modules to perform feature selection processing at two different scales, which does not constitute a limitation on this application. Optionally, the server may also include more feature selection processing modules, such as a third feature selection processing module, a fourth feature selection processing module, etc., to implement feature selection operations at more scales, and this application does not impose any restrictions on this. Optionally, after verification in the actual model training process, when there are three feature selection processing modules, a better model training effect can be obtained, and this application does not impose any restrictions on this.

[0080] For steps S341-S345, the server adopts the initial traffic prediction model to perform a first scale feature selection process on each spliced feature sample data in the spliced feature sample data set to obtain a first feature sample data set, and then perform a residual connection process on each spliced feature sample data in the spliced feature sample data set with each first feature sample data in the first feature sample data set to obtain a first residual feature sample data set, and perform a second scale feature selection process on each first residual feature sample data in the first residual feature sample data set to obtain a second feature sample data set, and then perform a residual connection process on each first residual feature sample data in the first residual feature sample data set with each second feature sample data in the second feature sample data set to obtain a second residual feature sample data set, so that each second residual feature sample data in the second residual feature sample data set can be fully connected to obtain a restored feature sample data set, and then a more accurate reference traffic prediction model can be trained based on the restored feature sample data set in subsequent steps.

[0081] It can be understood that the server uses the initial traffic prediction model to perform feature recovery on the replacement feature sample data in the replacement feature sample data set, and can obtain a restored feature sample data set (i.e., predicted feature data). The server can further reduce the difference between each restored feature sample data in the restored feature sample data set and each replacement feature label (real feature data) in the replacement feature label subset, thereby pre-training the initial traffic prediction model to obtain a reference traffic prediction model that better meets the needs. This application does not impose any restrictions on this.

[0082] For steps S31-S35, the server adopts an initial traffic prediction model to obtain a replacement feature label subset, a numerical replacement feature sample data subset and a category replacement feature sample data subset from the replacement feature sample data set, each replacement feature label in the replacement feature label subset is used to indicate the original feature corresponding to each replacement feature sample data, and each category replacement feature sample data in the category replacement feature sample data subset is embedded to obtain an embedded replacement feature sample data set, and each embedded replacement feature sample data in the embedded replacement feature sample data set is spliced with each numerical replacement feature sample data in the numerical replacement feature sample data subset to obtain a spliced feature sample data set, and each spliced feature sample data in the spliced feature sample data set is feature restored to obtain a restored feature sample data set, and then the initial traffic prediction model is pre-trained in the direction of reducing the difference between each restored feature sample data in the restored feature sample data set and each replacement feature label in the replacement feature label subset to obtain a more accurate reference traffic prediction model, which is conducive to obtaining a more accurate target traffic prediction model based on the reference traffic prediction model in subsequent steps.

[0083] S40: Input each labeled sample data in the labeled sample data set into the reference traffic prediction model to adjust model parameters to obtain a target traffic prediction model.

[0084] The target traffic prediction model can be understood as the traffic prediction model obtained by fine-tuning the model parameters of the reference traffic prediction model using the labeled sample data set. Optionally, the server can preprocess the labeled sample data in the labeled sample data set to further perform self-supervised training on the reference traffic prediction model using the preprocessed labeled sample data set, thereby obtaining the target traffic prediction model.

[0085] It should be noted that the labeled sample data set may include a labeled label subset, which may include one or more labeled labels. The labeled labels can be used to indicate the traffic usage information corresponding to each labeled sample data. This application does not impose any restrictions on this.

[0086] It should be understood that inputting each labeled sample data in the labeled sample data set into the reference traffic prediction model to adjust the model parameters and obtain the target traffic prediction model refers to the process of fine-tuning the model parameters of the pre-trained model. In particular, step S40, that is, inputting each labeled sample data in the labeled sample data set into the reference traffic prediction model to adjust the model parameters and obtain the target traffic prediction model, includes the following steps:

[0087] S41: using a reference traffic prediction model, performing feature enhancement processing on each labeled sample data in the labeled sample data set to obtain an enhanced feature sample data set;

[0088] S42: Using a reference traffic prediction model, perform traffic prediction on each enhanced feature sample data in the enhanced feature sample data set to obtain a reference traffic prediction result set;

[0089] S43: Based on the target loss function, the reference traffic prediction model is adjusted in the direction of reducing the difference between each reference traffic prediction result in the reference traffic prediction result set and each labeled label in the labeled label subset to obtain a target traffic prediction model.

[0090] The enhanced feature sample data set may include one or more enhanced feature sample data, which may be used to indicate the sample data obtained after the labeled sample data is subjected to feature enhancement processing. Optionally, the server may perform feature enhancement processing on the labeled sample data to make the features of the labeled sample data more prominent, thereby enabling the features to be learned to be acquired more quickly and accurately during the model training process. This application does not impose any restrictions on this.

[0091] Optionally, the server can copy the labeled sample data set to form the copied labeled sample data set as an index queue. Each sample data in the index queue can be understood as an index. The server can perform similarity calculation on each labeled sample data in the index queue, such as through cosine similarity calculation, to find the most similar m (m≥1) indexes from the index queue, and average the labels of the labeled sample data in the m indexes whose similarity is greater than a preset threshold to obtain new feature data (i.e., the above-mentioned enhanced feature sample data set), and then input the new feature data into the reference traffic prediction model for further fine-tuning of the model parameters.

[0092] Optionally, if there is no labeled sample data with a similarity greater than a preset threshold value in the m indexes, the server can use the average number of labels of the labeled sample data set as new feature data (i.e., the above-mentioned enhanced feature sample data set) to input the new feature data into the reference traffic prediction model for further fine-tuning of the model parameters. This application does not impose any restrictions on this. Optionally, after verification in the actual model training process, when the above-mentioned m is 7, a better model training effect can be obtained. This application does not impose any restrictions on this. Optionally, after verification in the actual model training process, when the above-mentioned preset threshold is 0.65, a better model training effect can be obtained. This application does not impose any restrictions on this.

[0093] The reference traffic prediction result set may include one or more reference traffic prediction results, which can be used to indicate the traffic prediction results obtained after traffic prediction is performed on enhanced feature sample data. The target loss function can be used to evaluate the performance of the reference traffic prediction model during the training process. By minimizing the value of the target loss function, the server can continuously adjust the parameters of the reference traffic prediction model to make the predicted traffic usage data closer to the actual traffic usage data, thereby improving the accuracy and generalization ability of the model.

[0094] Optionally, the objective loss function used in this application can be found in the following formula:

[0095]

[0096] in, represents the target loss function, y represents the labeled label in the labeled label subset, represents the reference traffic prediction result in the reference traffic prediction result set; n represents the number of labeled sample data contained in the labeled sample data set, i represents the index of the labeled sample data in the labeled sample data set; f(x) represents processing through the standard restriction function, x represents a temporary variable; y i Indicates the labeled label corresponding to the i-th labeled sample data in the labeled sample data set, represents the i-th reference traffic prediction result in the reference traffic prediction result set; e represents a natural constant; μ represents the mean corresponding to the labeled labels contained in the labeled sample data set, and σ represents the standard deviation corresponding to the labeled labels contained in the labeled sample data set.

[0097] The target loss function provided in the embodiment of the present application can obtain prior knowledge through a historical business data set, such as the mean μ corresponding to the labeled labels contained in the labeled sample data set, and the standard deviation σ corresponding to the labeled labels contained in the labeled sample data set, so that a stable output sequence can be obtained based on the prior knowledge, thereby enabling the model to quickly find the correct output range during the training process, which is conducive to the rapid convergence of the model to achieve efficient and accurate training. This application does not impose any restrictions on this.

[0098] It can be understood that the server uses the reference traffic prediction model to perform traffic prediction on the enhanced feature sample data in the enhanced feature sample data set based on the target loss function, and can obtain a reference traffic prediction result set (i.e., predicted traffic usage data). The server can further reduce the difference between each reference traffic prediction result in the reference traffic prediction result set and each labeled label in the labeled label subset (actual traffic usage data), thereby adjusting the model parameters of the reference traffic prediction model to obtain a more accurate and more demand-oriented target traffic prediction model. This application does not impose any restrictions on this.

[0099] For steps S41-S43, the server adopts a reference traffic prediction model to perform feature enhancement processing on each labeled sample data in the labeled sample data set to obtain an enhanced feature sample data set, and to perform traffic prediction on each enhanced feature sample data in the enhanced feature sample data set to obtain a reference traffic prediction result set, so that based on the target loss function, the reference traffic prediction model can be adjusted in the direction of reducing the difference between each reference traffic prediction result in the reference traffic prediction result set and each labeled label in the labeled label subset to obtain a target traffic prediction model, which is conducive to achieving more accurate traffic prediction based on the target traffic prediction model in subsequent steps, so as to effectively improve the accuracy of the traffic prediction results.

[0100] S50: Inputting the current service data set into the target traffic prediction model to perform traffic prediction and obtain a target traffic prediction result.

[0101] The target traffic prediction result may be used to indicate the traffic prediction result obtained after using the target traffic prediction model to perform traffic prediction on the service data in the current service data set. It is understood that performing traffic prediction using an accurate trained target traffic prediction model may result in a more accurate target traffic prediction result, and this application does not impose any limitation on this.

[0102] By inputting the current service data set into the trained target traffic prediction model, the server can accurately predict future traffic usage based on the current service data. Optionally, the server uses the target traffic prediction model to perform traffic prediction on each current service data in the current service data set. For details, see the aforementioned detailed description of the server using the initial traffic prediction model to process each historical service data in the historical service data set (i.e., the unlabeled sample data set and the labeled sample data set). This application will not elaborate on this further.

[0103] It can be seen that in the above scheme, the server obtains an unlabeled sample data set, a labeled sample data set and a current business data set to perform feature replacement processing on each unlabeled sample data in the unlabeled sample data set to obtain a replacement feature sample data set, so that each replacement feature sample data in the replacement feature sample data set can be input into the initial traffic prediction model for model pre-training to obtain a reference traffic prediction model, and each labeled sample data in the labeled sample data set can be input into the reference traffic prediction model for model parameter adjustment processing to obtain a more accurate target traffic prediction model, and then the current business data set can be input into the target traffic prediction model for traffic prediction to obtain a more accurate target traffic prediction result, which is conducive to obtaining a more accurate traffic prediction result.

[0104] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0105] In one embodiment, a traffic prediction device based on big data is provided, which corresponds one-to-one to the traffic prediction method based on big data in the above embodiment. Figure 3 As shown, the traffic prediction device based on big data includes an acquisition module 101, a first processing module 102, a second processing module 103, a third processing module 104 and a fourth processing module 105. The functional modules are described in detail as follows:

[0106] Acquisition module 101, used to acquire an unlabeled sample data set, a labeled sample data set and a current business data set;

[0107] A first processing module 102 is configured to perform feature replacement processing on each unlabeled sample data in the unlabeled sample data set to obtain a replacement feature sample data set;

[0108] The second processing module 103 is configured to input each replacement feature sample data in the replacement feature sample data set into an initial traffic prediction model for model pre-training to obtain a reference traffic prediction model;

[0109] The third processing module 104 is configured to input each labeled sample data in the labeled sample data set into the reference traffic prediction model to adjust the model parameters and obtain a target traffic prediction model;

[0110] The fourth processing module 105 is configured to input the current service data set into the target traffic prediction model to perform traffic prediction and obtain a target traffic prediction result.

[0111] In one embodiment, the second processing module 103 is specifically configured to use an initial traffic prediction model to obtain a replacement feature label subset, a numerical replacement feature sample data subset, and a category replacement feature sample data subset from the replacement feature sample data set, wherein each replacement feature label in the replacement feature label subset is used to indicate the original feature corresponding to each replacement feature sample data; use the initial traffic prediction model to embed each category replacement feature sample data in the category replacement feature sample data subset to obtain an embedded replacement feature sample data set; use the initial traffic prediction model to splice each embedded replacement feature sample data in the embedded replacement feature sample data set with each numerical replacement feature sample data in the numerical replacement feature sample data subset to obtain a spliced feature sample data set; use the initial traffic prediction model to perform feature recovery processing on each spliced feature sample data in the spliced feature sample data set to obtain a restored feature sample data set; and pre-train the initial traffic prediction model in a direction of reducing the difference between each restored feature sample data in the restored feature sample data set and each replacement feature label in the replacement feature label subset to obtain a reference traffic prediction model.

[0112] In one embodiment, the second processing module 103 is specifically used to use the initial traffic prediction model to perform a first scale feature selection process on each splicing feature sample data in the splicing feature sample data set to obtain a first feature sample data set; use the initial traffic prediction model to perform a residual connection process on each splicing feature sample data in the splicing feature sample data set and each first feature sample data in the first feature sample data set to obtain a first residual feature sample data set; use the initial traffic prediction model to perform a second scale feature selection process on each first residual feature sample data in the first residual feature sample data set to obtain a second feature sample data set; use the initial traffic prediction model to perform a residual connection process on each first residual feature sample data in the first residual feature sample data set and each second feature sample data in the second feature sample data set to obtain a second residual feature sample data set; use the initial traffic prediction model to perform a full connection process on each second residual feature sample data in the second residual feature sample data set to obtain a restored feature sample data set.

[0113] In one embodiment, the second processing module 103 is specifically used to use the initial traffic prediction model to extract the first-scale attention features of each spliced feature sample data in the spliced feature sample data set to obtain a first attention feature sample data set; use the initial traffic prediction model to obtain attention feature sample data that meets the preset conditions from the first attention feature sample data set to obtain a first target attention feature sample data subset; use the initial traffic prediction model to perform dimensionality increase mapping processing on each first target attention feature sample data in the first target attention feature sample data subset to obtain a first high-dimensional feature sample data set; use the initial traffic prediction model to activate each first high-dimensional feature sample data in the first high-dimensional feature sample data set to obtain a first activated feature sample data set; use the initial traffic prediction model to perform dimensionality reduction mapping processing on each first activated feature sample data in the first activated feature sample data set to obtain a first low-dimensional feature sample data set; use the initial traffic prediction model to perform residual connection processing on each spliced feature sample data in the spliced feature sample data set and each low-dimensional feature sample data in the first low-dimensional feature sample data set to obtain a first feature sample data set.

[0114] In one embodiment, the third processing module 104 is specifically configured to use a reference traffic prediction model to perform feature enhancement processing on each labeled sample data in the labeled sample data set to obtain an enhanced feature sample data set; use the reference traffic prediction model to perform traffic prediction on each enhanced feature sample data in the enhanced feature sample data set to obtain a reference traffic prediction result set; and based on a target loss function, adjust the model parameters of the reference traffic prediction model in a direction of reducing the difference between each reference traffic prediction result in the reference traffic prediction result set and each labeled label in the labeled label subset to obtain a target traffic prediction model;

[0115] The formula of the objective loss function is as follows:

[0116]

[0117] in, Represents the target loss function, y represents the labeled label, represents the reference traffic prediction result; n represents the number of labeled sample data contained in the labeled sample data set, i represents the index of the labeled sample data in the labeled sample data set; f(x) represents processing through the standard restriction function, x represents a temporary variable; y i Indicates the labeled label corresponding to the i-th labeled sample data in the labeled sample data set, represents the i-th reference traffic prediction result in the reference traffic prediction result set; e represents a natural constant; μ represents the mean corresponding to the labeled labels contained in the labeled sample data set, and σ represents the standard deviation corresponding to the labeled labels contained in the labeled sample data set.

[0118] The present invention provides a traffic prediction device based on big data, which obtains an unlabeled sample data set, a labeled sample data set and a current business data set, and performs feature replacement processing on each unlabeled sample data in the unlabeled sample data set to obtain a replacement feature sample data set, so that each replacement feature sample data in the replacement feature sample data set can be input into an initial traffic prediction model for model pre-training to obtain a reference traffic prediction model, and each labeled sample data in the labeled sample data set is input into the reference traffic prediction model for model parameter adjustment processing to obtain a more accurate target traffic prediction model, and then the current business data set can be input into the target traffic prediction model for traffic prediction to obtain a more accurate target traffic prediction result, which is conducive to obtaining a more accurate traffic prediction result.

[0119] For the specific definition of the traffic prediction device based on big data, please refer to the definition of the traffic prediction method based on big data above, which will not be repeated here. The various modules in the above-mentioned traffic prediction device based on big data can be implemented in whole or in part through software, hardware, and their combination. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0120] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 4 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client via a network connection. When the computer program is executed by the processor, it realizes the functions or steps on the server side of a traffic prediction method based on big data.

[0121] In one embodiment, a computer device is provided. The computer device may be a client, and its internal structure diagram may be as follows: Figure 5As shown. The computer device includes a processor, memory, network interface, display screen, and input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the functions or steps on the client side of a traffic prediction method based on big data.

[0122] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the following steps are performed:

[0123] Obtaining an unlabeled sample data set, a labeled sample data set, and a current business data set;

[0124] Perform feature replacement processing on each unlabeled sample data in the unlabeled sample data set to obtain a replacement feature sample data set;

[0125] Input each replacement feature sample data in the replacement feature sample data set into the initial traffic prediction model for model pre-training to obtain a reference traffic prediction model;

[0126] Input each labeled sample data in the labeled sample data set into the reference flow prediction model to adjust the model parameters to obtain the target flow prediction model;

[0127] The current business data set is input into the target traffic prediction model to perform traffic prediction and obtain the target traffic prediction result.

[0128] The present invention provides a computer device, which obtains an unlabeled sample data set, a labeled sample data set and a current business data set, and performs feature replacement processing on each unlabeled sample data in the unlabeled sample data set to obtain a replacement feature sample data set, so that each replacement feature sample data in the replacement feature sample data set can be input into an initial traffic prediction model for model pre-training to obtain a reference traffic prediction model, and each labeled sample data in the labeled sample data set can be input into the reference traffic prediction model for model parameter adjustment processing to obtain a more accurate target traffic prediction model, and then the current business data set can be input into the target traffic prediction model for traffic prediction to obtain a more accurate target traffic prediction result, which is conducive to obtaining a more accurate traffic prediction result.

[0129] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0130] Obtaining an unlabeled sample data set, a labeled sample data set, and a current business data set;

[0131] Perform feature replacement processing on each unlabeled sample data in the unlabeled sample data set to obtain a replacement feature sample data set;

[0132] Input each replacement feature sample data in the replacement feature sample data set into the initial traffic prediction model for model pre-training to obtain a reference traffic prediction model;

[0133] Input each labeled sample data in the labeled sample data set into the reference flow prediction model to adjust the model parameters to obtain the target flow prediction model;

[0134] The current business data set is input into the target traffic prediction model to perform traffic prediction and obtain the target traffic prediction result.

[0135] The present invention provides a computer-readable storage medium, which obtains an unlabeled sample data set, a labeled sample data set and a current business data set, and performs feature replacement processing on each unlabeled sample data in the unlabeled sample data set to obtain a replacement feature sample data set, so that each replacement feature sample data in the replacement feature sample data set can be input into an initial traffic prediction model for model pre-training to obtain a reference traffic prediction model, and each labeled sample data in the labeled sample data set is input into the reference traffic prediction model for model parameter adjustment processing to obtain a more accurate target traffic prediction model, and then the current business data set can be input into the target traffic prediction model for traffic prediction to obtain a more accurate target traffic prediction result, which is conducive to obtaining a more accurate traffic prediction result.

[0136] It should be noted that the above functions or steps that can be implemented by the computer-readable storage medium or computer device can be found in the relevant descriptions of the server side and the client side in the aforementioned method embodiment. To avoid repetition, they will not be described one by one here.

[0137] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0138] Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0139] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A traffic prediction method based on big data, characterized in that: The traffic prediction method based on big data includes: Obtaining an unlabeled sample data set, a labeled sample data set, and a current business data set; Performing feature replacement processing on each unlabeled sample data in the unlabeled sample data set to obtain a replaced feature sample data set; Using the initial traffic prediction model, obtaining a replacement feature label subset, a value replacement feature sample data subset, and a category replacement feature sample data subset from the replacement feature sample data set, wherein each replacement feature label in the replacement feature label subset is used to indicate the original feature corresponding to each replacement feature sample data; Using the initial traffic prediction model, embedding each category replacement feature sample data in the category replacement feature sample data subset to obtain an embedded replacement feature sample data set; Using the initial traffic prediction model, each embedded replacement feature sample data in the embedded replacement feature sample data set is spliced with each value-replaced feature sample data in the value-replaced feature sample data subset to obtain a spliced feature sample data set; Using the initial traffic prediction model, performing feature recovery processing on each spliced feature sample data in the spliced feature sample data set to obtain a restored feature sample data set; Pre-training the initial traffic prediction model in a direction of reducing the difference between each restored feature sample data in the restored feature sample data set and each replacement feature label in the replacement feature label subset to obtain a reference traffic prediction model; Inputting each labeled sample data in the labeled sample data set into the reference traffic prediction model to perform model parameter adjustment processing to obtain a target traffic prediction model; The current service data set is input into the target traffic prediction model to perform traffic prediction to obtain a target traffic prediction result.

2. The traffic prediction method based on big data according to claim 1, characterized in that: The initial traffic prediction model is used to perform feature recovery processing on each spliced feature sample data in the spliced feature sample data set to obtain a restored feature sample data set, including: Using the initial traffic prediction model, performing a first scale feature selection process on each splicing feature sample data in the splicing feature sample data set to obtain a first feature sample data set; Using the initial traffic prediction model, performing residual connection processing on each spliced feature sample data in the spliced feature sample data set and each first feature sample data in the first feature sample data set to obtain a first residual feature sample data set; Using the initial traffic prediction model, performing a second scale feature selection process on each first residual feature sample data in the first residual feature sample data set to obtain a second feature sample data set; Using the initial traffic prediction model, performing residual connection processing on each first residual feature sample data in the first residual feature sample data set and each second feature sample data in the second feature sample data set to obtain a second residual feature sample data set; The initial traffic prediction model is adopted to perform full connection processing on each second residual feature sample data in the second residual feature sample data set to obtain a restored feature sample data set.

3. The traffic prediction method based on big data according to claim 2 is characterized in that: The initial traffic prediction model is used to perform a first scale feature selection process on each spliced feature sample data in the spliced feature sample data set to obtain a first feature sample data set, including: Using the initial traffic prediction model, extracting first-scale attention features from each spliced feature sample data in the spliced feature sample data set to obtain a first attention feature sample data set; Using the initial traffic prediction model, acquiring attention feature sample data that meets preset conditions from the first attention feature sample data set to obtain a first target attention feature sample data subset; Using the initial traffic prediction model, performing dimensionality-increasing mapping processing on each first target attention feature sample data in the first target attention feature sample data subset to obtain a first high-dimensional feature sample data set; Using the initial traffic prediction model, activating each first high-dimensional feature sample data in the first high-dimensional feature sample data set to obtain a first activated feature sample data set; Using the initial traffic prediction model, performing dimensionality reduction mapping processing on each first activated feature sample data in the first activated feature sample data set to obtain a first low-dimensional feature sample data set; The initial traffic prediction model is adopted to perform residual connection processing on each spliced feature sample data in the spliced feature sample data set and each low-dimensional feature sample data in the first low-dimensional feature sample data set to obtain a first feature sample data set.

4. The traffic prediction method based on big data according to claim 3 is characterized in that: The labeled sample data set includes a labeled label subset, each labeled label in the labeled label subset is used to indicate traffic usage information corresponding to each labeled sample data; inputting each labeled sample data in the labeled sample data set into the reference traffic prediction model to perform model parameter adjustment processing to obtain a target traffic prediction model, including: Using a reference traffic prediction model, performing feature enhancement processing on each labeled sample data in the labeled sample data set to obtain an enhanced feature sample data set; Using a reference traffic prediction model, performing traffic prediction on each enhanced feature sample data in the enhanced feature sample data set to obtain a reference traffic prediction result set; Based on the target loss function, adjusting the model parameters of the reference traffic prediction model in a direction of reducing the difference between each reference traffic prediction result in the reference traffic prediction result set and each labeled label in the labeled label subset to obtain a target traffic prediction model; The formula of the objective loss function is as follows: in, Represents the target loss function, y represents the labeled label, represents the reference traffic prediction result; n represents the number of labeled sample data contained in the labeled sample data set, i represents the index of the labeled sample data in the labeled sample data set; f(x) represents processing through the standard restriction function, x represents a temporary variable; y i Indicates the labeled label corresponding to the i-th labeled sample data in the labeled sample data set, represents the i-th reference traffic prediction result in the reference traffic prediction result set; e represents a natural constant; μ represents the mean corresponding to the labeled labels contained in the labeled sample data set, and σ represents the standard deviation corresponding to the labeled labels contained in the labeled sample data set.

5. A traffic prediction device based on big data, characterized in that: The traffic prediction device based on big data includes: An acquisition module is used to acquire an unlabeled sample data set, a labeled sample data set, and a current business data set; A first processing module is configured to perform feature replacement processing on each unlabeled sample data in the unlabeled sample data set to obtain a replacement feature sample data set; a second processing module, configured to adopt an initial traffic prediction model to obtain a replacement feature label subset, a numerical replacement feature sample data subset, and a category replacement feature sample data subset from the replacement feature sample data set, wherein each replacement feature label in the replacement feature label subset is used to indicate the original feature corresponding to each replacement feature sample data; adopt the initial traffic prediction model to embed each category replacement feature sample data in the category replacement feature sample data subset to obtain an embedded replacement feature sample data set; adopt the initial traffic prediction model to splice each embedded replacement feature sample data in the embedded replacement feature sample data set with each numerical replacement feature sample data in the numerical replacement feature sample data subset to obtain a spliced feature sample data set; adopt the initial traffic prediction model to perform feature recovery processing on each spliced feature sample data in the spliced feature sample data set to obtain a restored feature sample data set; pre-train the initial traffic prediction model in a direction of reducing the difference between each restored feature sample data in the restored feature sample data set and each replacement feature label in the replacement feature label subset to obtain a reference traffic prediction model; A third processing module is configured to input each labeled sample data in the labeled sample data set into the reference traffic prediction model to perform model parameter adjustment processing to obtain a target traffic prediction model; The fourth processing module is used to input the current business data set into the target traffic prediction model to perform traffic prediction and obtain a target traffic prediction result.

6. The traffic prediction device based on big data according to claim 5, characterized in that: The second processing module is configured to use the initial traffic prediction model to perform feature recovery processing on each spliced feature sample data in the spliced feature sample data set to obtain a restored feature sample data set, specifically for: Using the initial traffic prediction model, performing a first scale feature selection process on each splicing feature sample data in the splicing feature sample data set to obtain a first feature sample data set; Using the initial traffic prediction model, performing residual connection processing on each spliced feature sample data in the spliced feature sample data set and each first feature sample data in the first feature sample data set to obtain a first residual feature sample data set; Using the initial traffic prediction model, performing a second scale feature selection process on each first residual feature sample data in the first residual feature sample data set to obtain a second feature sample data set; Using the initial traffic prediction model, performing residual connection processing on each first residual feature sample data in the first residual feature sample data set and each second feature sample data in the second feature sample data set to obtain a second residual feature sample data set; The initial traffic prediction model is adopted to perform full connection processing on each second residual feature sample data in the second residual feature sample data set to obtain a restored feature sample data set.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the traffic prediction method based on big data as described in any one of claims 1 to 4 is implemented.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the traffic prediction method based on big data as described in any one of claims 1 to 4 is implemented.

Images