Network element load prediction method, apparatus and device, and readable storage medium

By combining the time series prediction model and the traditional timing prediction model, the spatial and temporal dimension influence between network elements is combined, and the problem of insufficient efficiency and accuracy of network element load prediction is solved, and efficient network element load prediction and wireless network planning are achieved. .

CN119031397APending Publication Date: 2024-11-26CHINA MOBILE COMM LTD RES INST +2
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310589989.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-24
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

The prior art is difficult to effectively predict network element load, especially in intelligent wireless network planning. The lack of effective prediction methods leads to insufficient network planning efficiency and accuracy.

Method used

The fusion method of the time series prediction model and the traditional timing prediction model is adopted to predict the time series data of multi-network multi-load indicators through the time series prediction model, and the results of the traditional timing prediction model are fused to obtain the prediction results of the network element load. The method includes using attention mechanism to determine the spatial and temporal dimension impacts between network elements, improving prediction accuracy and generalization capabilities.

Benefits of technology

It significantly reduces the training and inference time of network element load prediction tasks, improves prediction efficiency and accuracy, corrects the lag and smoothing problems in long-sequence prediction, enhances the interpretability of the algorithm, supports intelligent wireless network planning, and saves Human and electricity costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119031397A_ABST
    Figure CN119031397A_ABST
Patent Text Reader

Abstract

The embodiment of the invention provides a network element load prediction method, device and equipment and a readable storage medium, and the method comprises the steps: obtaining a first prediction result through a time series prediction model, and the time series prediction model is used for the multi-load index time series data prediction of a plurality of network elements; obtaining a second prediction result through a traditional time sequence prediction model; and fusing the first prediction result and the second prediction result to obtain a prediction result of the network element load.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The embodiments of the present application relate to the field of communication technology, and specifically to a network element load prediction method, device, equipment and readable storage medium. Background Art

[0002] Intelligent wireless network planning refers to the application of intelligent technology to the process of network planning, which mainly includes two major directions: intelligent demand forecasting and intelligent planning scheme formulation. In the demand forecasting scenario, it is necessary to use intelligent artificial intelligence (AI) technology to build the relationship between network and business demand and make trend judgments based on network, user, and business development related data to achieve demand forecasting for network, business, and user development. For example, for all network elements in the region, the air interface uplink traffic, air interface downlink traffic, downlink physical resource block (PRB) average utilization and other indicators are forecasted in the medium and long term. At present, how to predict the network element load is an urgent problem to be solved. Summary of the invention

[0003] The embodiments of the present application provide a network element load prediction method, device, equipment and readable storage medium to solve the problem of how to predict the network element load.

[0004] In a first aspect, a network element load prediction method is provided, comprising:

[0005] Obtaining a first prediction result through a time series prediction model, wherein the time series prediction model is used for predicting time series data of multiple load indicators of multiple network elements;

[0006] A second prediction result is obtained through a traditional time series prediction model;

[0007] The first prediction result and the second prediction result are integrated to obtain a prediction result of the network element load.

[0008] Optionally, the method further includes:

[0009] Wireless network planning is performed according to the prediction result of the network element load.

[0010] Optionally, obtaining a first prediction result through a time series prediction model includes:

[0011] A first prediction result is obtained through a time series prediction model and an attention mechanism, wherein the attention structure is used to determine the influence of other network elements on the spatial dimension and / or temporal dimension of the current network element at each moment.

[0012] Optionally, the time series prediction model includes an encoder module, the input of the encoder module is a four-dimensional matrix, and the four-dimensional matrix includes: a first dimension, the first dimension is the batch size during model training and inference, a second dimension, the second dimension is the number of network elements processed by the model during model training and inference, a third dimension, the third dimension is the historical sequence length of the model input, and a fourth dimension, the fourth dimension is the encoding dimension of the model features.

[0013] Optionally, the time series prediction model includes a decoder module, the input of the decoder module is a four-dimensional matrix, and the four-dimensional matrix includes: a first dimension, the first dimension is the batch size during model training and inference, a second dimension, the second dimension is the number of network elements processed by the model during model training and inference, a third dimension, the third dimension is the historical sequence length and prediction sequence length of the model input, and a fourth dimension, the fourth dimension is the encoding dimension of the model features.

[0014] Optionally, the input of the time series prediction model includes the following four dimensions: feature coding, position coding, time granularity coding, and holiday or working day coding.

[0015] Optionally, the traditional time series prediction model includes a historical average prediction model and / or a periodic factor prediction model.

[0016] In a second aspect, a network element load prediction device is provided, comprising:

[0017] A first prediction module, used to obtain a first prediction result through a time series prediction model, wherein the time series prediction model is used to predict time series data of multiple load indicators of multiple network elements;

[0018] A second prediction module, used to obtain a second prediction result through a traditional time series prediction model;

[0019] A fusion module is used to fuse the first prediction result and the second prediction result to obtain a prediction result of the network element load.

[0020] In a third aspect, a communication device is provided, comprising a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the method described in the first aspect.

[0021] In a fourth aspect, a readable storage medium is provided, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the method described in the first aspect are implemented.

[0022] In an embodiment of the present application, a first prediction result is obtained through a time series prediction model, and the time series prediction model is used to predict time series data of multiple load indicators of multiple network elements; a second prediction result is obtained through a traditional time series prediction model; the first prediction result and the second prediction result are fused to obtain a prediction result of the network element load, thereby realizing the prediction of the network element load. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0024] Figure 1 is a flow chart of a network element load prediction method provided in an embodiment of the present application;

[0025] Figure 2 This is a schematic diagram of network element load prediction provided by an embodiment of the present application;

[0026] Figure 3a and Figure 3b is a schematic diagram of the attention calculation process provided by an embodiment of the present application;

[0027] Figure 4a , Figure 4b and Figure 4c It is a schematic diagram of the prediction results of uplink traffic, downlink traffic and PRB utilization of the time series prediction model and the traditional time series prediction model for network element 1 provided in an embodiment of the present application;

[0028] Figure 5a , Figure 5b and Figure 5c It is a schematic diagram of the prediction results of uplink traffic, downlink traffic and PRB utilization of the time series prediction model and the traditional time series prediction model for network element 2 provided in an embodiment of the present application;

[0029] Figure 6a , Figure 6b and Figure 6c It is a schematic diagram of the fusion effect of the network element 1 model;

[0030] Figure 7a , Figure 7b and Figure 7c It is a schematic diagram of the fusion effect of network element 2 model;

[0031] Figure 8 is a schematic diagram of a network element load prediction device provided in an embodiment of the present application;

[0032] Fig. 9It is a schematic diagram of a communication device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0034] The term "comprise" and any variation thereof in the specification and claims of the present application are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units need not be limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. In addition, the use of "and / or" in the specification and claims to indicate at least one of the connected objects, such as A and / or B, means that A alone, B alone, and A and B exist in three cases.

[0035] In the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0036] See also Figure 2 , an embodiment of the present application provides a network element load prediction method, and the specific steps include: step 201, step 202 and step 203.

[0037] Step 201: obtaining a first prediction result through a time series prediction model, wherein the time series prediction model is used for predicting time series data of multiple load indicators of multiple network elements;

[0038] It can be understood that the time series prediction model in this embodiment can improve the existing time series prediction network structure from single network element time series prediction to multi-network element and multi-load indicator time series data prediction, which can greatly reduce the training and reasoning time on massive network element load prediction tasks. In a single GPU (32GB) computing environment, it only takes about 10 seconds to complete the reasoning of 48 future data points of 20,000 network elements. In the existing time series prediction network structure, only one network element can be inferred at a time, that is, 20,000 model reasoning calls are required to complete the above tasks, which consumes a lot of time. Optionally, the time series prediction model in this embodiment can be an improved Informer model. Figures 4a to 7c In the example, it is represented as Informer*.

[0039] Step 202: Obtain a second prediction result through a traditional time series prediction model;

[0040] Optionally, the traditional time series prediction model includes a historical average prediction model and / or a periodic factor prediction model, wherein the time corresponding to the input of the historical average prediction model may be relevant data of the previous day, previous week and / or previous 2 weeks.

[0041] Optionally, the second forecast result of the historical average forecast model is Figures 4a to 7c The second prediction result of the periodic factor prediction model is Figures 4a to 7c It is represented as Period_factior in.

[0042] Step 203: The first prediction result and the second prediction result are integrated to obtain a prediction result of the network element load.

[0043] For example, the first prediction result and the second prediction result may include, but are not limited to, prediction results related to uplink traffic, downlink traffic and / or PRB utilization.

[0044] In this embodiment, the prediction results of the traditional time series prediction model and the time series prediction model are fused, so that the lag problem and smoothing problem in the long sequence prediction task in the prior art can be corrected.

[0045] It should be noted that the traditional time series prediction model has high requirements on the periodicity and stability of historical data and is prone to overfitting. The time series prediction model in this embodiment is too smooth compared to real data in long sequence prediction tasks. In step 203, the first prediction result and the second prediction result are fused to fully learn the three-dimensional influence relationship between each network element and each indicator, which not only improves the prediction efficiency and accuracy, but also improves the generalization ability and interpretability of the algorithm. The prediction result of the fused network element load (fusion_result) can be referred to Figures 6a to 7c.

[0046] In one embodiment of the present application, the method further includes:

[0047] Wireless network planning is performed according to the prediction result of the network element load.

[0048] In this embodiment, long-term regional traffic can be predicted to help achieve 5G base station site selection, planning and expansion, saving labor costs; predict large-scale base station user numbers, traffic and other indicators, and realize intelligent shutdown and wake-up of wireless base stations, saving base station power consumption.

[0049] Optionally, the wireless network planning includes but is not limited to determining at least one of the scale of network elements, the location of site establishment, and configuration parameters of network elements.

[0050] In one embodiment of the present application, a first prediction result is obtained by using a time series prediction model, including:

[0051] A first prediction result is obtained through a time series prediction model and an attention mechanism, wherein the attention structure is used to determine the influence of other network elements on the spatial dimension and / or temporal dimension of the current network element at each moment.

[0052] In this embodiment, the calculation of attention in the spatial and temporal dimensions is implemented, focusing on the impact of other network elements on the current network element at each moment, dynamically modeling the impact relationship between network elements, and improving the accuracy of prediction.

[0053] In one embodiment of the present application, the time series prediction model includes an encoder module, the input of the encoder module is a four-dimensional matrix, and the four-dimensional matrix includes: a first dimension, the first dimension is the batch size during model training and inference, a second dimension, the second dimension is the number of network elements processed by the model during model training and inference, a third dimension, the third dimension is the historical sequence length of the model input, and a fourth dimension, the fourth dimension is the encoding dimension of the model features.

[0054] In one embodiment of the present application, the time series prediction model includes a decoder module, the input of the decoder module is a four-dimensional matrix, and the four-dimensional matrix includes: a first dimension, the first dimension is the batch size during model training and inference, a second dimension, the second dimension is the number of network elements processed by the model during model training and inference, a third dimension, the third dimension is the historical sequence length and the predicted sequence length of the model input, and a fourth dimension, the fourth dimension is the encoding dimension of the model features.

[0055] In one embodiment of the present application, the input of the time series prediction model includes the following four dimensions: feature coding, position coding, time granularity coding, and holiday or workday coding.

[0056] In this embodiment, multi-network element time series prediction can be implemented, and traditional time series prediction algorithms can be integrated to improve the prediction accuracy, generalization and interpretability of long sequence prediction tasks.

[0057] The following is an introduction using the time series prediction model as an improved Informer network (for example, represented by Informer*) as an example.

[0058] The improved Informer network is used for multi-network element time series prediction, which greatly reduces the training and reasoning time on the massive network element load prediction task. It uses attention to capture the impact of network element indicators in the time and space dimensions, and integrates with traditional time series prediction algorithms, such as Figure 2 shown.

[0059] The time series prediction model in the figure is improved based on the Informer network, and the fusion module fuses the first prediction result of the time series prediction model with the second prediction results of the two traditional time series prediction models.

[0060] Input of the Encoder module in the time series forecasting model , is a four-dimensional matrix. The first dimension B is the batch size during model training and inference. The second dimension N is the number of network elements processed by the model during training and inference, for example, 1000. The third dimension his is the length of the historical sequence input to the model. Assuming that the time granularity of the network element indicator data is 1 hour and the inference is based on the historical 7-day data, his is 168. The fourth dimension d is the encoding dimension of the model features, for example, 512.

[0061] Decoder module input in time series forecasting model , is a four-dimensional matrix. The difference from the Encoder input matrix dimension is in the third dimension. his is the length of the input historical sequence, and pred represents the length of the predicted sequence. For example, if the indicator for the next 7 days is predicted, pred is 168. In the Decoder input, the sequence to be predicted uses 0 as a placeholder.

[0062] Figure 3a and Figure 3b is the attention calculation process, a 1 to a TFor the input sequence, we get the corresponding q, k, v vectors after linear mapping, which are called query vector, key value vector and value vector respectively. For the first element in the sequence, we use q 1 Multiply each key value vector and the result is processed by softmax operation. arrive , represents the element a 1 The degree of mutual influence between the input sequence and each element in the sequence is the attention score. The attention score and the value vector of each element are weighted to obtain the first element a of the input sequence of the module. 1 The output of 1 , which takes into account the characteristics of each element in the entire sequence.

[0063] Figure 3a and Figure 3b The calculation process is expressed in matrix form as follows: Q, K, and V are all 4-dimensional matrices, namely [B, N, L, D], where B represents the batch size, N represents the number of network elements, L represents the sequence length, and D represents the feature encoding dimension.

[0064]

[0065] The time series prediction model calculates the attention of the time dimension. The Q and K matrices perform matrix operations on the third and fourth dimensions, and the output dimension is [B, N, L, L]. A softmax operation is performed on the last dimension of the matrix to obtain the attention matrix A on the time dimension, which represents the correlation between each moment and other moments in the load indicator time series of each network element.

[0066] The time series prediction model calculates the attention of the spatial dimension. The Q and K matrices perform matrix operations on the second and fourth dimensions, and the output dimension is [B, N, N, L], which represents the correlation between each network element and other network elements at each moment. After the softmax operation, the spatial attention score matrix A is obtained, which is multiplied by the V matrix, and the result dimension is [B, N, L, D].

[0067] The input features of the model include the following four dimensions: (1) feature encoding, that is, encoding the network element load index; (2) position encoding. When the time series is input into the attention module for calculation, the front and back position relationship in the sequence will be lost. Therefore, it is necessary to add the position encoding of the relative position of the time series to the model input; (3) time encoding at hourly granularity; (4) holiday and weekday encoding. For long-series time series prediction problems, especially for network element load index data, there will be significantly different distribution patterns on weekdays and holidays.

[0068] In the task of predicting network element load indicators, the prediction results of the time series prediction model show a certain degree of lag and smoothness problems. The lag problem is that when the prediction results and the real indicators are visualized in the same image, it can be observed that the numerical changes of the prediction results lag slightly behind the real data. The smoothness problem is that the visualization curve of the prediction results is relatively smooth, and compared with the real indicator data, it lacks the details of ups and downs, but the overall trend is consistent with the real data.

[0069] In this embodiment, the prediction results of the time series prediction model are integrated with the prediction results of at least two traditional time series prediction algorithms, such as the historical average method (HA) and the period factor method (period factor). Because the traditional time series prediction algorithm only infers the future sequence based on the historical indicator data in the short term, it cannot capture nonlinear characteristics and the prediction results are often overfitted. Therefore, this embodiment mainly uses the time series prediction model, supplemented by HA and period factor, and integrates the prediction results of the model.

[0070] Figure 4a , Figure 4b and Figure 4c ,as well as Figure 5a , Figure 5b and Figure 5c The following are the prediction results of the time series prediction model (informer*), HA and periodic factor method for uplink traffic, downlink traffic and PRB utilization on two different network elements. The visualization curve trends of the three algorithms are consistent, the time series prediction model is smoother, and the other two algorithms have rich fluctuation details, but they are overfitted on the test data.

[0071] The inference results of the above three algorithms are weighted and fused, and the prediction result = αinformer* + βhistorical average + γperiodic factor, where α+β+γ=1. For example, 0.7<α<1, that is, Informer* is dominant in model fusion. The fusion effect is shown in Figure 6a , Figure 6b and Figure 6c ,as well as Figure 7a , Figure 7b and Figure 7c .

[0072] See also Figure 8 The embodiment of the present application provides a network element load prediction device, the device 800 comprising:

[0073] A first prediction module 801 is used to obtain a first prediction result through a time series prediction model, wherein the time series prediction model is used to predict time series data of multiple load indicators of multiple network elements;

[0074] The second prediction module 802 is used to obtain a second prediction result through a traditional time series prediction model;

[0075] The fusion module 803 is used to fuse the first prediction result and the second prediction result to obtain a prediction result of the network element load.

[0076] In one embodiment of the present application, the device further comprises:

[0077] The planning module is used to perform wireless network planning according to the prediction result of the network element load.

[0078] In one embodiment of the present application, a first prediction result is obtained by using a time series prediction model, including:

[0079] A first prediction result is obtained through a time series prediction model and an attention mechanism, wherein the attention structure is used to determine the influence of other network elements on the spatial dimension and / or temporal dimension of the current network element at each moment.

[0080] In one embodiment of the present application, the time series prediction model includes an encoder module, the input of the encoder module is a four-dimensional matrix, and the four-dimensional matrix includes: a first dimension, the first dimension is the batch size during model training and inference, a second dimension, the second dimension is the number of network elements processed by the model during model training and inference, a third dimension, the third dimension is the historical sequence length of the model input, and a fourth dimension, the fourth dimension is the encoding dimension of the model features.

[0081] In one embodiment of the present application, the time series prediction model includes a decoder module, the input of the decoder module is a four-dimensional matrix, and the four-dimensional matrix includes: a first dimension, the first dimension is the batch size during model training and inference, a second dimension, the second dimension is the number of network elements processed by the model during model training and inference, a third dimension, the third dimension is the historical sequence length and the predicted sequence length of the model input, and a fourth dimension, the fourth dimension is the encoding dimension of the model features.

[0082] In one embodiment of the present application, the input of the time series prediction model includes the following four dimensions: feature coding, position coding, time granularity coding, and holiday or workday coding.

[0083] In one embodiment of the present application, the traditional time series prediction model includes a historical average prediction model and / or a periodic factor prediction model.

[0084] The device provided in the embodiment of the present application can achieve Figure 1 The various processes implemented by the method embodiment shown achieve the same technical effect and will not be described again here to avoid repetition.

[0085] like Fig. 9 As shown, the embodiment of the present application also provides a communication device 900, including a processor 901, a memory 902, a program or instruction stored in the memory 902 and executable on the processor 901, and the program or instruction is executed by the processor 901 to implement the above Figure 1 The various processes of the method embodiment can achieve the same technical effect. To avoid repetition, they will not be described here.

[0086] The embodiment of the present application also provides a readable storage medium on which a program or instruction is stored. When the program or instruction is executed by a processor, the above Figure 1 The various processes of the method embodiment shown can achieve the same technical effect, and will not be described again here to avoid repetition.

[0087] The processor is the processor in the terminal described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0088] The steps of the method or algorithm described in conjunction with the contents disclosed in this application can be implemented in hardware or by executing software instructions in a processor. The software instructions can be composed of corresponding software modules, and the software modules can be stored in RAM, flash memory, ROM, EPROM, EEPROM, registers, hard disks, mobile hard disks, read-only optical disks, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be carried in an ASIC. In addition, the ASIC can be carried in a core network interface device. Of course, the processor and the storage medium can also exist in the core network interface device as discrete components.

[0089] Those skilled in the art should be aware that in one or more of the above examples, the functions described in this application can be implemented using hardware, software, firmware, or any combination thereof. When implemented using software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any media that facilitates the transmission of a computer program from one place to another. The storage medium can be any available medium that a general or special-purpose computer can access.

[0090] The specific implementation methods described above further illustrate the purpose, technical solutions and beneficial effects of the present application in detail. It should be understood that the above description is only the specific implementation methods of the present application and is not intended to limit the scope of protection of the present application. Any modifications, equivalent substitutions, improvements, etc. made on the basis of the technical solutions of the present application should be included in the scope of protection of the present application.

[0091] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the embodiments of the present application may adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the embodiments of the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0092] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0093] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0094] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0095] Obviously, those skilled in the art can make various changes and modifications to the embodiments of the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the embodiments of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A network element load prediction method, characterized in that: include: Obtaining a first prediction result through a time series prediction model, wherein the time series prediction model is used for predicting time series data of multiple load indicators of multiple network elements; Using the traditional time series prediction model, a second prediction result is obtained; The first prediction result and the second prediction result are integrated to obtain a prediction result of the network element load.

2. The method according to claim 1, characterized in that The method further comprises: Wireless network planning is performed according to the prediction result of the network element load.

3. The method according to claim 1, characterized in that Through the time series prediction model, the first prediction result is obtained, including: A first prediction result is obtained through a time series prediction model and an attention mechanism, wherein the attention structure is used to determine the influence of other network elements on the spatial dimension and / or temporal dimension of the current network element at each moment.

4. The method according to claim 1, characterized in that The time series prediction model includes an encoder module, the input of the encoder module is a four-dimensional matrix, and the four-dimensional matrix includes: a first dimension, the first dimension is the batch size during model training and inference, a second dimension, the second dimension is the number of network elements processed by the model during model training and inference, a third dimension, the third dimension is the historical sequence length of the model input, and a fourth dimension, the fourth dimension is the encoding dimension of the model features.

5. The method according to claim 1 or 4, characterized in that: The time series prediction model includes a decoder module, the input of the decoder module is a four-dimensional matrix, and the four-dimensional matrix includes: a first dimension, the first dimension is the batch size during model training and inference, a second dimension, the second dimension is the number of network elements processed by the model during model training and inference, a third dimension, the third dimension is the historical sequence length and the predicted sequence length of the model input, and a fourth dimension, the fourth dimension is the encoding dimension of the model features.

6. The method according to claim 1, characterized in that The input of the time series prediction model includes the following four dimensions: feature coding, position coding, time granularity coding, and holiday or working day coding.

7. The method according to claim 1, characterized in that The traditional time series prediction model includes a historical average prediction model and / or a periodic factor prediction model.

8. A network element load prediction device, characterized in that: include: A first prediction module, used to obtain a first prediction result through a time series prediction model, wherein the time series prediction model is used to predict time series data of multiple load indicators of multiple network elements; A second prediction module, used to obtain a second prediction result through a traditional time series prediction model; A fusion module is used to fuse the first prediction result and the second prediction result to obtain a prediction result of the network element load.

9. A communication device, characterized in that: The method comprises a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction implements the steps of the method according to any one of claims 1 to 7 when executed by the processor.

10. A readable storage medium, characterized in that: The readable storage medium stores a program or an instruction, and when the program or the instruction is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.