Network Load Prediction Method, Device, Electronic Device and Medium Based on Federated Learning

Abstract

The present application discloses a network load prediction method, apparatus, electronic device and medium based on federated learning. By applying the technical solution of the present application, a network slice user can use a trusted third-party device to receive the aggregation and transmission of model parameters in other network slice users under the federated learning architecture. Thus, based on the global model parameters, the training of the prediction model is carried out to obtain the trained target prediction model. Furthermore, it achieves the purpose of multi-party collaborative training of the slice load prediction model by using the federated learning framework, realizing slice-level distributed heterogeneous load prediction on the premise of ensuring user privacy and data non-sharing. At the same time, it also avoids the problem in the related art that the traffic prediction performance of the network slice is insufficient, which easily affects the progress of user services.

Description

Technical Field

[0001] The present application relates to data processing technologies, and in particular, to a network load prediction method, apparatus, electronic device, and medium based on federated learning. Background Art

[0002] Network Slicing (NS) is one of the core technologies enabling vertical industries in 5G / B5G communication systems, which refers to virtualizing the resources of a unified physical network infrastructure and abstracting them into multiple end-to-end (E2E) logical networks.

[0003] In related technologies, in order to realize the vision of automated management and orchestration of network slices, a running network slice needs to have the ability to perceive and predict in real time and actively respond to user requirements. In a wireless network, network load often has spatio-temporal correlation and can be predicted and perceived in advance. Among them, existing load prediction technologies are usually implemented through a centralized controller. For example, the controller collects global user data and network status information, centrally trains a global ML model, and performs prediction analysis to realize real-time perception of the future load status of the slice.

[0004] However, in the scenario of network slicing, a slice network may be deployed on multiple base stations, and its historical load data is geographically dispersed. Moreover, network slices are isolated from each other, and user data is privacy-sensitive. It is difficult to establish a centralized machine learning model to analyze and optimize the traffic prediction performance of network slices. As a result, problems that affect the progress of business occur. Summary of the Invention

[0005] Embodiments of the present application provide a network load prediction method, apparatus, electronic device, and medium based on federated learning, which are used to solve the problem in related technologies that the traffic prediction performance of network slices is insufficient and easily affects the progress of user services.

[0006] Among them, according to one aspect of the embodiments of the present application, a network load prediction method based on federated learning is provided, which is applied to target network slice users and includes:

[0007] Receiving a public key sent by a third-party device, where the public key is used to encrypt transmission data;

[0008] Using the public key to encrypt local initial model parameters and then transmitting them to the third-party device, where the initial model parameters are obtained by the target network slice training its own deployed initial prediction model using sample data;

[0009] Receive the global model parameters sent by the third-party device, and based on the global model parameters, obtain a trained target prediction model, where the global model parameters are the model parameters obtained by the third-party device by aggregating a plurality of the initial model parameters;

[0010] Use the target prediction model to predict the load of the target network slice in a future time period, and send the load prediction result to the target base station associated with the target network slice.

[0011] Optionally, in another embodiment based on the above method of the present application, before encrypting the local initial model parameters with the public key and transmitting them to the third-party device, it further includes:

[0012] Receive the sample load data encrypted with the public key sent by the third-party device;

[0013] Decrypt the encrypted sample load data to obtain the sample load data, and then use the sample load data to train the initial prediction model to obtain the initial model parameters.

[0014] Optionally, in another embodiment based on the above method of the present application, the training the initial prediction model with the sample load data to obtain the initial model parameters includes:

[0015] Use the sample load data to train the initial prediction model to obtain the gradient value and loss value corresponding to the initial prediction model;

[0016] Take the gradient value and the loss value as the initial model parameters.

[0017] Optionally, in another embodiment based on the above method of the present application, after taking the gradient value and the loss value as the initial model parameters, it further includes:

[0018] Use the public key to encrypt the gradient value and the loss value, and then send the encrypted gradient value and loss value to the third-party device;

[0019] Receive the total gradient value parameter sent by the third-party device, where the total gradient value parameter is the model parameter obtained by the third-party device by aggregating a plurality of the gradient values and loss values;

[0020] Based on the total gradient value parameter, obtain the trained target prediction model.

[0021] Optionally, in another embodiment based on the above method of the present application, the obtaining the trained target prediction model based on the global model parameters includes:

[0022] Use the global model parameters to perform a preset model accuracy test on the initial movement prediction model;

[0023] If it is determined that the initial movement prediction model meets the preset model accuracy, use the initial movement prediction model as the target movement prediction model; or,

[0024] If it is determined that the initial movement prediction model does not meet the preset model accuracy, use the global model parameters to train the initial movement prediction model to obtain the target movement prediction model.

[0025] Optionally, in another embodiment based on the method of the present application, after sending the load prediction result to the target base station associated with the target network slice, it further includes:

[0026] Receive a resource allocation instruction sent by the target base station, where the resource allocation instruction is generated by the target base station based on the load prediction result.

[0027] Wherein, according to another aspect of the embodiments of the present application, a network load prediction device based on federated learning is provided, which is applied to a target network slice user and includes:

[0028] A receiving module, configured to receive a public key sent by a third-party device, where the public key is used for encrypting transmission data;

[0029] An encryption module, configured to use the public key to encrypt the local initial model parameters and then transmit them to the third-party device, where the initial model parameters are obtained by the target network slice training its own deployed initial prediction model using sample data;

[0030] A generation module, configured to receive the global model parameters sent by the third-party device and obtain a trained target prediction model based on the global model parameters, where the global model parameters are model parameters obtained by the third-party device aggregating multiple initial model parameters;

[0031] A prediction module, configured to use the target prediction model to predict the load of the target network slice in a future time period and send the load prediction result to the target base station associated with the target network slice.

[0032] According to another aspect of the embodiments of the present application, an electronic device is provided, including:

[0033] A memory for storing executable instructions; and

[0034] A display for operating with the memory to execute the executable instructions to complete the operation of any one of the above-described federated learning-based network load prediction methods.

[0035] According to another aspect of the embodiments of the present application, there is provided a computer-readable storage medium for storing computer-readable instructions, which, when executed, perform the operations of any one of the above-described federated learning-based network load prediction methods.

[0036] In the present application, a public key sent by a third-party device can be received, and the public key is used to encrypt the transmitted data; the initial model parameters locally are encrypted with the public key and then transmitted to the third-party device, and the initial model parameters are obtained by the target network slice training its own deployed initial prediction model using sample data; receive the global model parameters sent by the third-party device, and based on the global model parameters, obtain the trained target prediction model, where the global model parameters are the model parameters obtained by the third-party device summarizing multiple initial model parameters; use the target prediction model to predict the load of the target network slice in a future time period, and send the load prediction result to the target base station associated with the target network slice. By applying the technical solution of the present application, the network slice user can use a trusted third-party device to receive the summary and transmission of the model parameters in other network slice users under the federated learning architecture. Thus, based on the global model parameters, the prediction model is trained to obtain the trained target prediction model. Furthermore, it achieves the purpose of multi-party collaborative training of the slice load prediction model using the federated learning framework, and can realize slice-level distributed heterogeneous load prediction while ensuring user privacy and data non-sharing. At the same time, it also avoids the problem in the related art that the traffic prediction performance of the network slice is insufficient, which easily affects the progress of user services.

[0037] The technical solution of the present application will be further described in detail below with reference to the drawings and embodiments. Description of the Drawings

[0038] The drawings constituting a part of the specification depict the embodiments of the present application and, together with the description, are used to explain the principles of the present application.

[0039] Referring to the drawings, the present application can be more clearly understood according to the following detailed description, where:

[0040] Figure 1 Shows a schematic diagram of a federated learning-based network load prediction method provided by an embodiment of the present application;

[0041] Figure 2 Shows a schematic diagram of a federated learning-based network load prediction system architecture provided by an embodiment of the present application;

[0042] Figure 3 shows a schematic flow chart of a network load prediction method provided by an embodiment of the present application based on federated learning;

[0043] Figure 4 shows a schematic structural diagram of an electronic device provided by an embodiment of the present application;

[0044] Figure 5 shows a schematic structural diagram of an electronic device provided by an embodiment of the present application;

[0045] Figure 6 shows a schematic diagram of a storage medium provided by an embodiment of the present application. Detailed implementation manners

[0046] Now, various exemplary embodiments of the present application will be described in detail with reference to the accompanying drawings. It should be noted that: unless otherwise specifically stated, the relative arrangements of components and steps, numerical expressions, and values set forth in these embodiments do not limit the scope of the present application.

[0047] At the same time, it should be understood that, for the sake of description, the dimensions of the various parts shown in the drawings are not drawn in actual proportional relationships.

[0048] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way a limitation on the present application, its application, or its use.

[0049] Technologies, methods, and devices known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, such technologies, methods, and devices should be regarded as part of the specification.

[0050] It should be noted that: like reference numerals and letters denote like items in the following drawings, and thus, once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.

[0051] In addition, the technical solutions between various embodiments of the present application can be combined with each other, but it must be based on the fact that those of ordinary skill in the art can implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the present application.

[0052] It should be noted that all directional indications (such as up, down, left, right, front, back...) in the embodiments of the present application are only used to explain the relative positional relationship and movement conditions between components in a specific posture (as shown in the drawings). If the specific posture changes, the directional indications will also change accordingly.

[0053] Next, in combination withFigures 1 - 3 To describe a method for network load prediction based on federated learning according to an exemplary embodiment of the present application. It should be noted that the following application scenarios are only shown for the convenience of understanding the spirit and principle of the present application, and the embodiments of the present application are not limited in this regard. On the contrary, the embodiments of the present application can be applied to any applicable scenario.

[0054] The present application also proposes a method, device, electronic device and medium for network load prediction based on federated learning.

[0055] Figure 1 Schematically shows a flowchart of a method for network load prediction based on federated learning according to an embodiment of the present application. As Figure 1 shown, this method is applied to target network slice users and includes:

[0056] S101, receiving a public key sent by a third-party device, where the public key is used to encrypt transmission data.

[0057] S102, using the public key to encrypt the local initial model parameters and then transmitting them to the third-party device. The initial model parameters are obtained by the target network slice training its own deployed initial prediction model using sample data.

[0058] S103, receiving the global model parameters sent by the third-party device, and obtaining a trained target prediction model based on the global model parameters. The global model parameters are the model parameters obtained by the third-party device aggregating multiple initial model parameters.

[0059] S104, using the target prediction model to predict the load of the target network slice for a future time period, and sending the load prediction result to the target base station associated with the target network slice.

[0060] In the related art, network slicing (NS) is one of the core technologies for 5G / B5G communication systems to empower vertical industries, which refers to virtualizing the resources of the unified physical network infrastructure and abstracting them into multiple end-to-end (E2E) logical networks.

[0061] Furthermore, these logical networks are isolated from each other and independently serve a specific business scenario, so as to meet the user's customized and diversified quality of service (QoS) requirements. Network slicing is realized relying on SDN / NFV technology and is a collection of a group of virtual network functions (VNFs) and the resources they possess. Different VNFs can implement a specific network function.

[0062] Among them, a network slice is composed of multiple VNFs, thus forming a complete logical network to realize customized services. In a network slice, it can be divided into three parts: radio network (AN) sub-slice, bearer network (TN) sub-slice and core (CN) network sub-slice. In the 5G era, ITU formally defines three typical application scenarios of network slicing: enhanced mobile broadband (eMBB), massive machine type communication (mMTC), and ultra-reliable and low latency communication (URLLC).

[0063] At present, network slicing can basically realize the cross-domain integration of end-to-end business processes and provide customized services to users. The technical goal of the next stage is to realize cross-domain automated management, orchestration and configuration of network slicing and enhance the end-to-end automation capability of network slicing under the premise of guaranteeing the service level agreement (SLA). This is inseparable from the support of load sensing technology driven by artificial intelligence (AI) and big data.

[0064] Furthermore, for the use of network slicing to implement AI-based load perception and traffic prediction technology, in order to realize the vision of automated management and orchestration of network slicing, the running slices need to have the ability to perceive and predict in real time and actively respond to user needs. In wireless networks, network loads often have temporal and spatial correlations and can be predicted and perceived in advance. Due to differences in business types and user groups, the traffic patterns of different slices are heterogeneous (for example, different business types have different traffic peak periods, and some scenarios have stronger traffic bursts, etc.).

[0065] The essence of network traffic prediction is time series prediction. Traditional traffic prediction is mainly based on time series models. The specific parameters of the model are solved according to actual data, and finally the time series model with known parameters is used for time series prediction. Common time series models include Moving Average (MA), Auto Regressive (AR), Auto Regressive Moving Average (ARMA), and Auto Regressive Integrated Moving Average (ARIMA).

[0066] In view of the above problems, the present application proposes a network load prediction method based on federated learning. The idea is that network slice users use a trusted third-party device to receive the aggregation and transmission of model parameters in other network slice users under the federated learning architecture. Then, based on the global model parameters, the training of the prediction model is carried out to obtain the trained target prediction model. Thus, under the premise of ensuring user privacy and data non-sharing, multi-party collaborative training of the slice load prediction model is achieved by using the federated learning framework, and distributed heterogeneous load prediction at the slice level can be realized. At the same time, it also avoids the problem in related technologies that the traffic prediction performance of network slices is insufficient, which is likely to affect the progress of user services.

[0067] Furthermore, in the 6G era, a large amount of user data will be distributed at the network edge, which requires the wireless network to have more ubiquitous and native intelligent capabilities. Specifically, AI is not only deployed in the centralized SDN controller that manages slices, but also exists in network nodes and user equipment (UE) in a distributed paradigm. In various distributed network architectures, federated learning (FL) is considered a potentially important solution for realizing 6G ubiquitous intelligence due to its advantages such as protecting privacy and overcoming data silos.

[0068] Among them, federated learning is essentially a distributed machine learning framework. The goal is to achieve joint modeling and improve the performance of the AI model on the basis of ensuring data privacy security and compliance. Under the FL framework, local clients use a large amount of local user data to train local models and upload the local model parameters to a centralized base station / edge server for aggregation, so as to obtain a global model, and multi-party collaborative training of the ML model can be realized under the premise of protecting user privacy and data non-sharing.

[0069] In one way, the prediction model in the embodiment of the present application can be a linear regression model (LinearRegression, LR), a support vector regression model (Support Vector Regression, SVR), a long short-term memory model (Long Short Time Memory, LSTM) neural network, etc.

[0070] As an example, the prediction model can be a long short-term memory model. It can be understood that the LSTM network is an improved recurrent neural network (RNN), which can solve the problem that the RNN cannot handle long-distance dependencies. It adds a state (called the cell state) to the hidden layer of the original RNN to save the long-term state, and introduces a gate structure to control the retention and discard of historical memory information, control the retention of highly relevant information, and delete less relevant content.

[0071] Therefore, LSTM has good memory performance and can effectively solve the gradient explosion and gradient vanishing problems of RNN. Network traffic prediction based on ML is mainly implemented through centralized network elements such as data centers, centralized controllers (such as SDN controllers) or intelligent network functions (such as Network Data Analytics Function (NWDAF)). The centralized control unit uniformly collects global user data and network status information, performs intelligent analysis, prediction and decision-making on the data, and realizes the perception of the future load status of the slice.

[0072] In one way, Figure 2 The figure shows the system architecture diagram of the network load prediction method based on federated learning. In the two-layer heterogeneous network scenario composed of macro and micro base stations, different network slices (composed of different service flows) share the same physical network infrastructure in one area. That is, one micro base station can carry multiple slices, or one slice can be deployed on multiple micro base stations.

[0073] Among them, users requesting the same service (slice) may access different micro base stations. Therefore, the historical traffic data in each slice is distributed on multiple micro base stations and is geographically dispersed. In order to collect distributed slice data and provide relatively reliable slice-level predictions with lower communication overhead and latency, in the solution of this application, inter-slice load prediction is implemented based on vertical federated learning.

[0074] Further, such as Figure 3 FIG. 1 is a flow chart of a network load prediction method based on federated learning proposed in this application, which includes:

[0075] Step 1: Perform sample encryption alignment between slice users participating in federated learning, and confirm the common users of both parties without disclosing their respective data, so as to combine the features of these users for modeling.

[0076] It should be noted that after determining the common user group, the initial prediction model needs to be trained using the data owned by each target slice user. In order to ensure the confidentiality of the data during the training process, it is necessary to use encrypted third-party equipment for encrypted training.

[0077] It should be noted that the target slice user in this application may be multiple slice users, and the multiple slice users need to be supported by one base station device.

[0078] Among them, slice users are the holders of user data in the slice network.

[0079] Step 2: The encryption third-party device sends the public key to each target slice user for encryption of subsequent transmission data.

[0080] It should be noted that the encrypted third-party device in this application can be a server or a server cluster. This application does not make any restrictions on this.

[0081] Step 3: The encrypted third-party device encrypts the sample payload data using the public key and sends the encrypted sample payload data to the target slice user.

[0082] Step 4: Each target slice user trains the initial prediction model using the decrypted sample payload data to obtain the gradient value and loss value corresponding to the initial prediction model, and encrypts and sends the gradient value and loss value to the encrypted third-party device.

[0083] Step 5: The encrypted third-party device aggregates the obtained gradient values and loss values sent by each target slice user and decrypts them.

[0084] Step 6: The encrypted third-party device sends the decrypted total gradient value parameter back to each target slice user.

[0085] Step 7: Each slice user updates its respective initial prediction model parameter according to the total gradient value parameter.

[0086] Step 8: Each slice user performs an accuracy test on the initial prediction model to determine whether the current model converges. There are two cases:

[0087] The first case: If it is determined that the initial mobile prediction model meets the preset model accuracy, the initial mobile prediction model is used as the target mobile prediction model.

[0088] The second case: If it is determined that the initial mobile prediction model does not meet the preset model accuracy, the initial mobile prediction model is trained using the global model parameter until the target mobile prediction model is obtained.

[0089] Step 9: Use the target prediction model to predict the load of the target network slice in the future time period and send the load prediction result to the target base station associated with the target network slice.

[0090] So far, the inter-slice load prediction based on vertical federated learning is completed.

[0091] Step 10: Receive the resource allocation instruction sent by the target base station, where the resource allocation instruction is generated by the target base station based on the load prediction result, so that the target base station executes the next decision process according to the prediction result and then notifies the corresponding target network slice.

[0092] In this application, a public key sent by a receiving third-party device can be used to encrypt transmission data; the initial model parameters locally are encrypted using the public key and then transmitted to the third-party device, where the initial model parameters are obtained by the target network slice training its locally deployed initial prediction model using sample data; receive the global model parameters sent by the third-party device, and based on the global model parameters, obtain the trained target prediction model, where the global model parameters are the model parameters obtained by the third-party device aggregating multiple initial model parameters; use the target prediction model to predict the load of the target network slice for a future time period, and send the load prediction result to the target base station associated with the target network slice. By applying the technical solution of this application, a network slice user can use a trusted third-party device to receive the aggregation and transmission of model parameters in other network slice users under a federated learning architecture. Thus, based on the global model parameters, the prediction model is trained to obtain the trained target prediction model. Furthermore, it achieves the purpose of multi-party collaborative training of a slice load prediction model using a federated learning framework, while ensuring user privacy and data non-sharing, and enables slice-level distributed heterogeneous load prediction. At the same time, it also avoids the problem in related technologies that the traffic prediction performance of the network slice is insufficient, which easily affects the progress of user services.

[0093] Optionally, in another embodiment based on the above method of this application, before encrypting the local initial model parameters using the public key and transmitting them to the third-party device, it further includes:

[0094] Receive the sample load data encrypted using the public key sent by the third-party device;

[0095] Decrypt the encrypted sample load data to obtain the sample load data, and then use the sample load data to train the initial prediction model to obtain the initial model parameters.

[0096] Optionally, in another embodiment based on the above method of this application, the using the sample load data to train the initial prediction model to obtain the initial model parameters includes:

[0097] Use the sample load data to train the initial prediction model to obtain the gradient value and loss value corresponding to the initial prediction model;

[0098] Take the gradient value and loss value as the initial model parameters.

[0099] Optionally, in another embodiment based on the above method of this application, after taking the gradient value and loss value as the initial model parameters, it further includes:

[0100] After encrypting the gradient value and the loss value by using the public key, send the encrypted gradient value and loss value to the third-party device;

[0101] Receive the total gradient value parameter sent by the third-party device, where the total gradient value parameter is a model parameter obtained by the third-party device by aggregating a plurality of the gradient values and loss values;

[0102] Based on the total gradient value parameter, obtain the trained target prediction model.

[0103] Optionally, in another embodiment based on the method of the present application above, the obtaining the trained target prediction model based on the global model parameter includes:

[0104] Use the global model parameter to perform a preset model accuracy test on the initial movement prediction model;

[0105] If it is determined that the initial movement prediction model meets the preset model accuracy, use the initial movement prediction model as the target movement prediction model; or,

[0106] If it is determined that the initial movement prediction model does not meet the preset model accuracy, train the initial movement prediction model by using the global model parameter to obtain the target movement prediction model.

[0107] Optionally, in another embodiment based on the method of the present application above, after sending the load prediction result to the target base station associated with the target network slice, further includes:

[0108] Receive the resource allocation instruction sent by the target base station, where the resource allocation instruction is generated by the target base station based on a decision made according to the load prediction result.

[0109] In one way, the technical solution proposed by the present application can provide a slice-level heterogeneous load prediction method based on federated learning. Thus, it can achieve multi-party collaborative training of the slice load prediction model by using the federated learning framework, while ensuring user privacy and data non-sharing, and can realize slice-level distributed heterogeneous load prediction, and improve the prediction performance.

[0110] Among them, since each slice user adopts the scheme of only uploading model parameters instead of original data, it can effectively reduce the communication overhead and delay caused by data exchange, and at the same time can effectively avoid the privacy sensitivity and data security problems caused by data sharing.

[0111] Optionally, in another implementation manner of the present application, as Figure 4As shown in the figure, the present application also provides a network load prediction device based on federated learning. It includes:

[0112] A receiving module 201, configured to receive a public key sent by a third-party device, where the public key is used to encrypt transmission data;

[0113] An encryption module 202, configured to use the public key to encrypt the local initial model parameters and then transmit them to the third-party device, where the initial model parameters are obtained by the target network slice training its own deployed initial prediction model using sample data;

[0114] A generating module 203, configured to receive the global model parameters sent by the third-party device and, based on the global model parameters, obtain a trained target prediction model, where the global model parameters are model parameters obtained by the third-party device summarizing multiple initial model parameters;

[0115] A prediction module 204, configured to use the target prediction model to predict the load of the target network slice for a future time period and send the load prediction result to a target base station associated with the target network slice.

[0116] By applying the technical solution of the present application, a network slice user can use a trusted third-party device to receive the summary and transmission of model parameters in other network slice users under the federated learning architecture. Thus, based on the global model parameters, the prediction model is trained to obtain a trained target prediction model. Furthermore, it achieves the purpose of multi-party collaborative training of the slice load prediction model using the federated learning framework, and can realize slice-level distributed heterogeneous load prediction while ensuring user privacy and data non-sharing. At the same time, it also avoids the problem in the related technology that the traffic prediction performance of the network slice is insufficient, which easily affects the progress of user services.

[0117] In another implementation manner of the present application, the steps that the receiving module 201 is configured to execute include:

[0118] Receive the encrypted sample load data sent by the third-party device using the public key;

[0119] Decrypt the encrypted sample load data to obtain the sample load data, and then use the sample load data to train the initial prediction model to obtain the initial model parameters.

[0120] In another implementation manner of the present application, the steps that the receiving module 201 is configured to execute include:

[0121] Train the initial prediction model using the sample load data to obtain the gradient value and loss value corresponding to the initial prediction model;

[0122] Use the gradient value and loss value as the initial model parameters.

[0123] In another implementation manner of the present application, the steps performed by the receiving module 201 include:

[0124] After encrypting the gradient value and the loss value using the public key, send the encrypted gradient value and loss value to the third-party device;

[0125] Receive the total gradient value parameter sent by the third-party device, where the total gradient value parameter is the model parameter obtained by the third-party device by aggregating multiple gradient values and loss values;

[0126] Based on the total gradient value parameter, obtain the trained target prediction model.

[0127] In another implementation manner of the present application, the steps performed by the receiving module 201 include:

[0128] Use the global model parameters to perform a preset model accuracy test on the initial mobile prediction model;

[0129] If it is determined that the initial mobile prediction model meets the preset model accuracy, use the initial mobile prediction model as the target mobile prediction model; or,

[0130] If it is determined that the initial mobile prediction model does not meet the preset model accuracy, train the initial mobile prediction model using the global model parameters to obtain the target mobile prediction model.

[0131] In another implementation manner of the present application, the steps performed by the receiving module 201 include:

[0132] Receive the resource allocation instruction sent by the target base station, where the resource allocation instruction is generated by the target base station based on the load prediction result.

[0133] An embodiment of the present application further provides an electronic device to execute the above network load prediction method based on federated learning. Please refer to Figure 5 , which shows a schematic diagram of an electronic device provided by some embodiments of the present application. As Figure 5As shown, the electronic device 3 includes: a processor 300, a memory 301, a bus 302, and a communication interface 303. The processor 300, the communication interface 303, and the memory 301 are connected through the bus 302. The memory 301 stores a computer program that can run on the processor 300. When the processor 300 runs the computer program, it executes the network load prediction method based on federated learning provided by any of the foregoing embodiments of the present application.

[0134] Among them, the memory 301 may include a high-speed random access memory (RAM: Random Access Memory), and may also include a non-volatile memory, such as at least one disk memory. Through at least one communication interface 303 (which can be wired or wireless), a communication connection is established between this device network element and at least one other network element. The Internet, wide area network, local area network, metropolitan area network, etc. can be used.

[0135] The bus 302 can be an ISA bus, a PCI bus, an EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. Among them, the memory 301 is used to store a program. After receiving an execution instruction, the processor 300 executes the program. The data recognition method disclosed in any of the foregoing embodiments of the present application can be applied to the processor 300 or implemented by the processor 300.

[0136] The processor 300 may be an integrated circuit chip with the ability to process signals. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in the processor 300 or the instructions in the form of software. The above-mentioned processor 300 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present application can be directly embodied as being executed and completed by the hardware decoding processor, or executed and completed by the combination of the hardware and software modules in the decoding processor. The software module may be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. This storage medium is located in the memory 301, and the processor 300 reads the information in the memory 301 and combines its hardware to complete the steps of the above method.

[0137] The electronic device provided by the embodiments of the present application and the method for predicting network load based on federated learning provided by the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run, or implemented by them.

[0138] The embodiments of the present application also provide a computer-readable storage medium corresponding to the method for data recognition provided in the foregoing embodiments. Please refer to Figure 6 , which shows that the computer-readable storage medium is an optical disc 40, on which a computer program (i.e., a program product) is stored. When the computer program is run by a processor, it will execute the method for predicting network load based on federated learning provided in any of the foregoing embodiments.

[0139] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory, or other optical and magnetic storage media, which will not be elaborated here one by one.

[0140] The computer-readable storage medium provided by the above embodiments of the present application and the method for data recognition provided by the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0141] It should be noted that:

[0142] In the specification provided herein, a large number of specific details are set forth. However, it is understood that the embodiments of the present application may be practiced without these specific details. In some instances, well-known structures and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0143] Similarly, it should be understood that in order to streamline the present application and assist in understanding one or more of the various inventive aspects, in the above description of the exemplary embodiments of the present application, the various features of the present application are sometimes grouped together into a single embodiment, figure, or description thereof. However, the disclosed method should not be construed as reflecting the following schematic: that the claimed present application requires more features than are expressly recited in each claim. Rather, as reflected in the following claims, the inventive aspects lie in less than all the features of the single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into the detailed description, with each claim standing on its own as a separate embodiment of the present application.

[0144] In addition, those skilled in the art will appreciate that although some of the embodiments described herein include certain features included in other embodiments but not others, the combination of features of different embodiments is within the scope of the present application and forms different embodiments. For example, in the following claims, any one of the claimed embodiments can be used in any combination.

[0145] The above is only the preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A network load prediction method based on federated learning, characterized in that, Applied to target network slice users, including: Receiving a public key sent by a third-party device, where the public key is used to encrypt transmission data; Using the public key to encrypt the local initial model parameters and then transmitting them to the third-party device, where the initial model parameters are obtained by the target network slice training its deployed initial prediction model using sample data; Receiving the global model parameters sent by the third-party device, and based on the global model parameters, obtaining a trained target prediction model, where the global model parameters are the model parameters obtained by the third-party device summarizing multiple initial model parameters; Using the target prediction model to predict the load of the target network slice for a future time period, and sending the load prediction result to a target base station associated with the target network slice; The obtaining the trained target prediction model based on the global model parameters includes: Using the global model parameters to perform a preset model accuracy test on the initial mobile prediction model; If it is determined that the initial mobile prediction model meets the preset model accuracy, using the initial mobile prediction model as the target mobile prediction model; If it is determined that the initial mobile prediction model does not meet the preset model accuracy, training the initial mobile prediction model using the global model parameters to obtain the target mobile prediction model.

2. The method according to claim 1, characterized in that, Before using the public key to encrypt the local initial model parameters and then transmitting them to the third-party device, it further includes: Receiving encrypted sample load data encrypted using the public key sent by the third-party device; Decrypting the encrypted sample load data to obtain the sample load data, and then using the sample load data to train the initial prediction model to obtain the initial model parameters.

3. The method according to claim 2, characterized in that, The training the initial prediction model using the sample load data to obtain the initial model parameters includes: Using the sample load data to train the initial prediction model to obtain the gradient value and loss value corresponding to the initial prediction model; Taking the gradient value and loss value as the initial model parameters.

4. The method according to claim 3, characterized in that, After taking the gradient value and loss value as the initial model parameters, it further includes: Using the public key to encrypt the gradient value and the loss value, and then sending the encrypted gradient value and loss value to the third-party device; Receiving the total gradient value parameter sent by the third-party device, where the total gradient value parameter is the model parameter obtained by the third-party device summarizing multiple gradient values and loss values; Based on the total gradient value parameter, obtaining the trained target prediction model.

5. The method according to claim 1, characterized in that, After sending the load prediction result to the target base station associated with the target network slice, it further includes: Receiving a resource allocation instruction sent by the target base station, where the resource allocation instruction is generated by the target base station making a decision based on the load prediction result.

6. A network load prediction device based on federated learning, characterized in that, Applied to target network slice users, including: A receiving module, configured to receive a public key sent by a third-party device, where the public key is used for encrypting transmission data; An encryption module, configured to use the public key to encrypt local initial model parameters and then transmit them to the third-party device, where the initial model parameters are obtained by the target network slice training an initial prediction model deployed by itself using sample data; A generation module, configured to receive the global model parameters sent by the third-party device and, based on the global model parameters, obtain a trained target prediction model, where the global model parameters are model parameters obtained by the third-party device aggregating a plurality of the initial model parameters; A prediction module, configured to use the target prediction model to perform a load prediction for the target network slice in a future time period and send the load prediction result to a target base station associated with the target network slice; The steps performed by the receiving module include: Using the global model parameters to perform a preset model accuracy test on an initial mobile prediction model; If it is determined that the initial mobile prediction model meets the preset model accuracy, using the initial mobile prediction model as the target mobile prediction model; If it is determined that the initial mobile prediction model does not meet the preset model accuracy, training the initial mobile prediction model using the global model parameters to obtain the target mobile prediction model.

7. An electronic device, characterized in that, Including: A memory, used for storing executable instructions; And, A processor, configured to execute the executable instructions with the memory to complete the operations of the network load prediction method based on federated learning according to any one of claims 1-5.

8. A computer-readable storage medium for storing computer-readable instructions, characterized in that, When the instructions are executed, the operations of the network load prediction method based on federated learning according to any one of claims 1-5 are performed.

Images