Network capacity prediction method and device, electronic equipment and storage medium

By using the capacity prediction model of graph convolutional units and gated cycle units in large venues, combining the correlation weights of fusion evaluation parameters and network management side performance parameters, the problem of low accuracy of short-term network capacity prediction is solved, and rapid and accurate network capacity prediction, supporting real-time optimization.

CN119071802AActive Publication Date: 2024-12-03CHINA MOBILE GROUP ZHEJIANG +2

Patent Information

Application Number
CN202411125952.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-16
Publication Date
2024-12-03
Estimated Expiration
2044-08-16

AI Technical Summary

Technical Problem

The accuracy of short-term network capacity prediction in large venues in the prior art is not high, and traditional methods cannot be quickly adjusted to adapt to the rapid changes in the business model, resulting in the accuracy of network capacity prediction results that need to be improved.

Method used

The capacity prediction model based on graph convolution unit and gated loop unit is adopted, and network capacity prediction is achieved by obtaining the correlation weight between the fusion evaluation parameters and the performance parameters on the network management side, and the connection between user trajectory changes and spatial changes is captured.

Benefits of technology

It improves the accuracy and efficiency of short-term network capacity prediction of large venues, can quickly capture time and space characteristics, provide more accurate network capacity prediction values, and supports real-time network optimization and adjustment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119071802A_ABST
    Figure CN119071802A_ABST
Patent Text Reader

Abstract

The invention provides a network capacity prediction method and apparatus, an electronic device and a storage medium. The method comprises the steps of obtaining fusion evaluation parameters of a to-be-predicted region; and inputting the fusion evaluation parameters and association weights between any network management side performance parameter in the fusion evaluation parameters and all fusion evaluation parameters except any network management side performance parameter into a capacity prediction model to obtain a network capacity prediction value. According to the method provided by the invention, the association weights between the fusion evaluation parameters and the performance parameters of any network management side and all the fusion evaluation parameters except the performance parameters of any network management side are input into the capacity prediction model, and the user track change and the space change are reflected through the association weights. And the capacity prediction model is indicated to capture the relation between the user track change and the space change in the fusion evaluation parameters, so that rapid and accurate network capacity prediction is realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a network capacity prediction method, device, electronic equipment and storage medium. Background Art

[0002] Compared with general areas, large venues have high density of people and higher business concurrency requirements. Therefore, the network capacity prediction problem of large venues has always been a hot topic for improving wireless network user performance. Traditional wireless network optimization cannot be adjusted quickly according to the rapid changes in business models, and there is a "lag". Therefore, the current common network capacity prediction methods are usually based on the Transformer and ARIMA models, as well as feature decomposition of historical data to achieve network capacity prediction.

[0003] However, the current common network service forecasts are mainly based on a single historical indicator, and some characteristics are biased, so the accuracy of the network capacity forecast results needs to be further improved. Summary of the invention

[0004] The present invention provides a network capacity prediction method, device, electronic device and storage medium, which are used to solve the defect that the accuracy of the network capacity prediction result needs to be further improved in the prior art for short-term network capacity prediction.

[0005] The present invention provides a network capacity prediction method, comprising: Obtain fusion evaluation parameters of the area to be predicted; Input the fusion evaluation parameter and the association weight between any network management side performance parameter in the fusion evaluation parameter and each fusion evaluation parameter other than the any network management side performance parameter into a capacity prediction model, so as to obtain a network capacity prediction value of the area to be predicted outputted by the capacity prediction model; The capacity prediction model is obtained by training an initial capacity prediction model based on sample fusion evaluation parameters.

[0006] According to a network capacity prediction method provided by the present invention, the capacity prediction model includes a graph convolution unit and a gated recurrent unit; The step of inputting the fusion evaluation parameter and the association weight between any network management side performance parameter in the fusion evaluation parameter and each fusion evaluation parameter other than the any network management side performance parameter into a capacity prediction model, and obtaining a network capacity prediction value of the area to be predicted outputted by the capacity prediction model, comprises: Extracting parameter features of the fusion evaluation parameters at the historical moment based on the graph convolution unit and the associated weights; Based on the gated recurrent unit and the self-attention mechanism, feature processing is performed on the parameter features of the historical moments to obtain the parameter features of the current moment; Based on the parameter characteristics at the current moment, the network capacity prediction value is obtained.

[0007] According to a network capacity prediction method provided by the present invention, the gated cycle unit includes an update gate, a reset gate, a candidate hidden gate, and an update hidden gate; The feature processing of the parameter features of the historical moments based on the gated recurrent unit and the self-attention mechanism to obtain the parameter features of the current moment includes: Based on the update gate and the first attention weight corresponding to the update gate, feature retaining the parameter feature of the historical moment to obtain a first retained feature; Based on the reset gate and the second attention weight corresponding to the reset gate, feature discarding is performed on the parameter feature of the historical moment to obtain a second retained feature; Based on the candidate hidden door and the third attention weight corresponding to the candidate hidden door, feature processing is performed on the parameter feature of the historical moment and the second retained feature to obtain a third retained feature; Based on the updated hidden gate, feature processing is performed on the first retained feature, the second retained feature, and the third retained feature to obtain the current moment parameter feature.

[0008] According to a network capacity prediction method provided by the present invention, obtaining the network capacity prediction value based on the parameter characteristics at the current moment includes: Based on the prediction function and the parameter characteristics at the current moment, an initial capacity prediction value is obtained; Performing network capacity prediction on the area to be predicted based on the network adjustment parameters of the area to be predicted to obtain a capacity adjustment prediction value; The initial capacity prediction value is corrected based on the capacity adjustment prediction value to obtain the network capacity prediction value.

[0009] According to a network capacity prediction method provided by the present invention, the network adjustment parameters at least include the number of system switching times between base stations and individual cell offset parameters.

[0010] According to a network capacity prediction method provided by the present invention, the fusion evaluation parameters also include base station side physical parameters and software measurement configuration parameters; The step of obtaining the association weight between any network management side performance parameter in the fusion evaluation parameter and each fusion evaluation parameter other than the any network management side performance parameter comprises: Based on the graph convolutional network, connection edges are calculated for any of the network management side performance parameters and each of the fusion evaluation parameters, so as to obtain an association weight between any of the network management side performance parameters and each of the fusion evaluation parameters except any of the network management side performance parameters.

[0011] The present invention also provides a network capacity prediction device, comprising: An acquisition unit, which acquires fusion evaluation parameters of the area to be predicted; The prediction unit inputs the fusion evaluation parameter and the association weight between any network management side performance parameter in the fusion evaluation parameter and each fusion evaluation parameter other than the any network management side performance parameter into a capacity prediction model, so as to obtain a network capacity prediction value of the area to be predicted outputted by the capacity prediction model; The capacity prediction model is obtained by training an initial capacity prediction model based on sample fusion evaluation parameters.

[0012] The present invention also provides an electronic 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 program, any of the network capacity prediction methods described above is implemented.

[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the network capacity prediction method as described in any one of the above is implemented.

[0014] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the network capacity prediction method described above is implemented.

[0015] The network capacity prediction method, device, electronic device and storage medium provided by the present invention input the association weights between the fusion evaluation parameter and any network management side performance parameter and each fusion evaluation parameter excluding any network management side performance parameter into the capacity prediction model, and instruct the capacity prediction model to capture the connection between the user trajectory change and the spatial change in the fusion evaluation parameter through the connection weight. This realizes fast and accurate network capacity prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are 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 work.

[0017] Figure 1This is one of the flow charts of the network capacity prediction method provided by the present invention; Figure 2 This is the second flow chart of the network capacity prediction method provided by the present invention; Figure 3 It is a comparative schematic diagram of feature reconstruction by fusing evaluation parameters provided by the present invention; Figure 4 It is a schematic diagram of a feature reconstruction relationship map obtained based on the maximum number of RRC users provided by the present invention; Figure 5 This is the third flow chart of the network capacity prediction method provided by the present invention; Figure 6 It is a schematic diagram of the relationship between the user flow trajectory and the physical characteristics provided by the present invention; Figure 7 It is a structural schematic diagram of a network capacity prediction device provided by the present invention; Figure 8 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] Currently, the short-term prediction methods for large venues mainly use two models: Transformer and ARIMA. Among them, Transformer is a widely used and versatile prediction model. The main processing method of the model is to optimize the self-attention mechanism and perform feature decomposition on the input sequence. This design enables Transformer to capture richer time dependencies with lower complexity when processing longer historical data. However, although these optimizations reduce some of the computational burden, the large parameter size of the Transformer model causes its computational cost to be significantly higher than that of a simple linear network model, and this increased computational burden does not significantly improve the model's prediction accuracy.

[0020] In addition, the basic idea of ​​the ARIMA model is to use the historical information of the data itself to predict the future. The label value at a point in time is affected by both the label value in the past period of time and the accidental events in the past period of time. That is to say, the label value of the ARIMA model fluctuates around the general trend of time, where the trend is affected by historical labels and the fluctuation is affected by accidental events in a period of time. The general trend itself is not necessarily stable, so this method is not effective in dealing with short-term predictions.

[0021] Therefore, the current network capacity prediction scheme does not have high accuracy in short-term prediction in large venues. To solve the above problem, the present invention provides a network capacity prediction method to solve the defect of low accuracy of short-term network prediction in large venues, so as to achieve accurate and efficient short-term network capacity prediction. Figure 1 FIG. 1 is one of the flow charts of the network capacity prediction method provided by the present invention, such as Figure 1 As shown, the method includes: Step 110, obtaining fusion evaluation parameters of the area to be predicted; Step 120, inputting the fusion evaluation parameter and the association weight between any network management side performance parameter in the fusion evaluation parameter and each fusion evaluation parameter excluding the any network management side performance parameter into a capacity prediction model, and obtaining a network capacity prediction value of the area to be predicted outputted by the capacity prediction model; The capacity prediction model is obtained by training an initial capacity prediction model based on sample fusion evaluation parameters.

[0022] Here, the area to be predicted refers to a large venue or other area with high traffic density and high business concurrency demand. The convergence evaluation parameters of the area to be predicted may include network management side performance parameters and other network parameters that do not change over time. The network management side performance parameters here may include the maximum number of RRC (Radio Resource Control) users, uplink and downlink PRB (Physical Resource Block) utilization, traffic, etc. It can be understood that the network management side performance parameters reflect parameters that change over time.

[0023] In addition, the network management side performance parameters in the fusion evaluation parameters can reflect the operating status of the network in the test area, and then reflect the actual carrying capacity and performance of the network, that is, the time characteristics can be reflected through the network management side performance parameters. It can be understood that when the area with a larger flow of people, the actual carrying capacity of the network is stronger; conversely, the actual carrying capacity of the network is relatively weak. In addition, other network parameters that do not change with time can reflect the physical parameter characteristics of the test area, such as location, coverage, signal strength, etc., and other fusion evaluation parameters can reflect spatial characteristics.

[0024] The capacity prediction model here can be constructed based on a graph convolutional network, and the initial capacity prediction model can be iteratively trained by fusion evaluation parameters of samples acquired in advance to optimize the model parameters of the initial capacity prediction model until an accurate capacity prediction model is obtained.

[0025] Specifically, when predicting the short-term network capacity of the area to be predicted, the fusion evaluation parameters of the area to be predicted at the historical moment can be obtained first. For example, the fusion evaluation parameters of each minute in the previous 10 minutes are obtained. Then, the fusion evaluation parameters and the association weights between any pre-acquired network management side performance parameters and each fusion evaluation parameter other than any network management side performance parameter can be input into the capacity prediction model. The relationship between the user trajectory and the spatial characteristics reflected by the association weights can help the capacity prediction model capture the features associated with the network capacity prediction value in the fusion evaluation parameters, and then output the network capacity prediction value of the area to be predicted through the capacity prediction model, so as to achieve accurate network capacity prediction. The network capacity prediction value here can be the network management side performance parameters at future moments, i.e., uplink PRB utilization, downlink PRB utilization, RRC maximum number of users, and traffic.

[0026] It should be noted that feature calculation is performed on the network management side performance parameters and other fusion evaluation parameters in the fusion evaluation parameters to achieve reconstruction of the association between time features and space features, and capture the association between time features and space features between the fusion evaluation parameters. That is, the association weight between any network management side performance parameter and each fusion evaluation parameter other than any network management side performance parameter can be used to reflect the degree of association between user trajectories and physical parameters in the area to be predicted. It can be understood that by capturing the relationship between user trajectories and physical parameters in the fusion evaluation parameters, network usage requirements at different times and locations can be more accurately predicted, so as to better optimize network performance.

[0027] It should also be noted that the association weights between any network management side performance parameter and each fusion evaluation parameter other than any network management side performance parameter are calculated in advance. Therefore, when the network capacity is predicted through the capacity prediction model, the calculation process of the association weights can be removed, which not only simplifies the model structure of the capacity prediction model, but also greatly improves the efficiency of network capacity prediction, thereby realizing short-term capacity prediction of large venue networks.

[0028] The method provided by the embodiment of the present invention inputs the association weights between the fusion evaluation parameter and any network management side performance parameter and each fusion evaluation parameter excluding any network management side performance parameter into the capacity prediction model, and instructs the capacity prediction model to capture the connection between the user trajectory change and the spatial change in the fusion evaluation parameter through the connection weight. This realizes fast and accurate network capacity prediction.

[0029] Based on any of the above embodiments, the capacity prediction model includes a graph convolution unit and a gated recurrent unit; The above step 120 includes: Extracting parameter features of the fusion evaluation parameters at the historical moment based on the graph convolution unit and the associated weights; Based on the gated recurrent unit and the self-attention mechanism, feature processing is performed on the parameter features of the historical moments to obtain the parameter features of the current moment; Based on the parameter characteristics at the current moment, the network capacity prediction value is obtained.

[0030] Here, the capacity prediction model may include an input layer, a graph convolution unit, a gated recurrent unit, and an output layer. Furthermore, the input layer, the graph convolution unit, the gated recurrent unit, and the output layer are connected in sequence. In addition, the graph convolution unit may be constructed based on a multi-layer GCN (Graph Convolutional Network) network. For example, the number of GCN layers may be set to 3 to reduce the amount of short-term computation, which can be used for real-time prediction and is suitable for short-term capacity prediction of large venue networks.

[0031] Specifically, first, the fusion evaluation parameters and each associated weight of the historical moment are input into the input layer. Then, the fusion evaluation parameters and each associated weight of the historical moment are transmitted to the graph convolution unit through the input layer. In the graph convolution unit, the relationship graph corresponding to the fusion evaluation parameters can be reconstructed by each associated weight. That is, taking each fusion evaluation parameter as a node, and using each associated weight as the connecting edge between the corresponding two nodes, feature reconstruction is performed to obtain a reconstructed relationship graph corresponding to the fusion evaluation parameter. Then, the reconstructed relationship graph is convolved based on the graph convolution unit to extract the node features of each node, that is, to obtain the parameter features. It can be understood that the parameter features here can reflect the connection between the user trajectory and the spatial features between the fusion evaluation parameters.

[0032] Furthermore, the parameter features are input into the gated recurrent unit, and based on the gated recurrent unit and the self-attention mechanism, the parameter features are processed so that the feature information related to the network capacity prediction is enhanced and receives more attention, so as to obtain the parameter features at the current moment. Finally, the parameter features at the current moment are input into the output layer, and the network capacity value is predicted based on the prediction function of the output layer and the parameter features at the current moment to obtain the predicted network capacity value.

[0033] Based on any of the above embodiments, the gated cycle unit includes an update gate, a reset gate, a candidate hidden gate, and an update hidden gate; The feature processing of the parameter features of the historical moments based on the gated recurrent unit and the self-attention mechanism to obtain the parameter features of the current moment includes: Based on the update gate and the first attention weight corresponding to the update gate, feature retaining the parameter feature of the historical moment to obtain a first retained feature; Based on the reset gate and the second attention weight corresponding to the reset gate, feature discarding is performed on the parameter feature of the historical moment to obtain a second retained feature; Based on the candidate hidden door and the third attention weight corresponding to the candidate hidden door, feature processing is performed on the parameter feature of the historical moment and the second retained feature to obtain a third retained feature; Based on the updated hidden gate, feature processing is performed on the first retained feature, the second retained feature, and the third retained feature to obtain the current moment parameter feature.

[0034] Specifically, the update gate is used to retain features, that is, based on the first attention weight corresponding to the update gate, the features that need to be retained at the next moment are selected from the parameter features of the previous moment to obtain the first retained features. The first retained features here can be calculated by the following formula, as shown below:

[0035] In the formula, represents the first retained feature at time t; Represents the sigmoid activation function; represents the first attention weight; express The final calculated value of the feature of a hidden layer at the time; Represents the parameter characteristics at time t.

[0036] The reset gate is used to discard features, that is, based on the second attention weight corresponding to the reset gate, the features that need to be discarded at the next moment are selected from the parameter features of the previous moment to obtain the second retained features. The second retained features here can be calculated by the following formula, as shown below:

[0037] In the formula, represents the second retained feature at time t.

[0038] The candidate hidden gate is used to represent the feature with a smaller weight. Based on the third attention weight corresponding to the candidate hidden gate, the parameter feature of the historical moment and the second retained feature are processed to obtain the third retained feature. The third retained feature here can be calculated by the following formula, as shown in the following formula:

[0039] In the formula, represents the third retained feature at time t; is the hyperbolic tangent function.

[0040] The updated hidden gate is used to perform feature processing on the first retained feature, the second retained feature, and the third retained feature to obtain the feature that needs to be retained, that is, the parameter feature at the current moment. The parameter feature at the current moment here can be calculated by the following formula, as shown in the following formula:

[0041] In the formula, Indicates the parameter characteristics at the current moment.

[0042] It should be noted that the gated recurrent unit controls the flow and forgetting of information through the update gate and reset gate, thereby achieving modeling of sequence data. The update gate determines how much of the current moment's information should come from the hidden state of the previous moment, while the reset gate controls the impact of past information on the current moment. This gating mechanism enables the gated recurrent unit to better capture important information in the sequence when processing feature sequence data, improving the accuracy of network capacity prediction. At the same time, it alleviates the gradient vanishing problem, making the model easier to train and optimize.

[0043] In practical applications, wireless network optimization and adjustment is a dynamic process. That is, load balancing between base stations usually gives priority to adjusting CIO parameters, and the number of handovers between base stations will also change significantly with the adjustment of CIO. Therefore, in order to further improve the accuracy of network capacity prediction, based on any of the above embodiments, the network capacity prediction value is obtained based on the parameter characteristics at the current moment, including: Based on the prediction function and the parameter characteristics at the current moment, an initial capacity prediction value is obtained; Performing network capacity prediction on the area to be predicted based on the network adjustment parameters of the area to be predicted to obtain a capacity adjustment prediction value; The initial capacity prediction value is corrected based on the capacity adjustment prediction value to obtain the network capacity prediction value.

[0044] Based on any of the above embodiments, the network adjustment parameters at least include the number of system switching times between base stations and individual cell offset parameters.

[0045] Specifically, the initial capacity prediction value can be calculated by inputting the parameter features at the current moment into the prediction function. Thus, the short-term capacity prediction of large venues can be completed according to the prediction function generated by the GCN model. The indicator changes in the next 5 minutes can be predicted based on the historical 10-minute performance indicators, supporting the 5-minute granularity time series feature prediction and user trajectory prediction.

[0046] It should be noted that the essence of time series prediction is to capture the time-varying patterns in the input data, and to fit these patterns by establishing mathematical models to minimize the error between the model prediction value and the actual observation value. The capacity prediction model here adopts the GCN graph convolutional network, which uses data-driven learning of graph structure to achieve automatic updating of the graph. By constructing an adjacency matrix, the hidden spatial dependencies are retained, and at the same time, invisible graph structures can be discovered from the data without the guidance of any prior knowledge. In the training process of the capacity prediction model, the core convolutional layer formula parameters and variables are obtained by training the GCN model. Therefore, in the actual application of capacity prediction, the prediction function can be expressed by the following formula, as shown below:

[0047] In the formula, represents the prediction function; Indicates the number of layers of the GCN neural network; represents the normal transformation matrix; represents the inverse transformation matrix; , , Respectively The weight of the layer feature; X represents the parameter feature of the input at the current moment.

[0048] Furthermore, the network capacity of the area to be predicted can be predicted by the network adjustment parameters of the area to be predicted, and the capacity adjustment prediction value can be obtained. Here, the network adjustment parameters of the area to be predicted can include the number of system switching between base stations and the individual offset parameters of the cell in the area to be measured at the historical moment, wherein the number of system switching between base stations can include the number of same-system switching, such as the number of switching between 4G and 4G base stations, or the number of switching between 5G and 5G base stations. And the number of switching between different systems, such as the number of switching between 4G base stations and 5G base stations. Therefore, based on the parameters of the network dynamic optimization adjustment at the historical moment, the network capacity of the area to be predicted can be predicted, and the capacity adjustment prediction value can be obtained. Furthermore, the initial capacity prediction value can be corrected based on the capacity adjustment prediction value. For example, the capacity adjustment prediction value and the initial capacity prediction value are added to obtain the final network capacity prediction value, so as to achieve a more accurate network capacity prediction.

[0049] In one embodiment, Figure 2 FIG. 2 is a flow chart of the network capacity prediction method provided by the present invention. Figure 2 As shown, the method includes: First, the performance data on the network management side, the physical information on the base station side, the configuration information on the software side, and the corresponding weights are input into the GRU gating unit, and the GRU gating unit is used to perform deep learning time series prediction to predict the initial capacity prediction value. At the same time, the network adjustment parameters of the area to be predicted are input into the model deviation correction module to obtain the capacity adjustment prediction value. Finally, the initial capacity prediction value is corrected based on the capacity adjustment prediction value to obtain the final network capacity prediction value.

[0050] The method provided by the embodiment of the present invention takes into account that in the process of short-term network capacity prediction, since wireless network optimization and adjustment is a dynamically changing process, a model correction unit is added after a preliminary network capacity prediction is performed based on the fusion evaluation parameters, and the network capacity after network optimization and adjustment is predicted by using the network optimization and adjustment parameters at the past moment, so that the initial capacity prediction value is corrected based on the capacity adjustment prediction value, which fully considers the impact of wireless network optimization and adjustment on the network capacity prediction result, and greatly improves the accuracy of short-term network capacity prediction.

[0051] Based on any of the above embodiments, the fusion evaluation parameters further include base station side physical parameters and software measurement configuration parameters; The step of obtaining the association weight between any network management side performance parameter in the fusion evaluation parameter and each fusion evaluation parameter other than the any network management side performance parameter comprises: Based on the graph convolutional network, connection edges are calculated for any of the network management side performance parameters and each of the fusion evaluation parameters, so as to obtain an association weight between any of the network management side performance parameters and each of the fusion evaluation parameters except any of the network management side performance parameters.

[0052] Here, the physical parameters on the base station side may include the base station height, antenna azimuth, antenna downtilt, station spacing, and RRU (Remote Radio Unit and Radio Remote Unit) capabilities. It should be noted that the physical parameters on the base station side jointly affect the layout and performance of the wireless communication network. For example, the base station height and antenna azimuth determine the coverage direction and range of the signal, the antenna downtilt is used to adjust the vertical coverage area of ​​the signal, and the station spacing affects the density and capacity of the network. It should be noted that compared with the existing technology, when predicting network capacity, its evaluation parameters no longer only consider dynamically changing performance indicators, but also physical site information. By reconstructing the time domain and spatial domain features, the shortcomings of traditional capacity prediction in multi-feature fusion are compensated, and the key to feature reconstruction is to use the node relationship of the GCN graph convolutional network to calculate the weight, thereby improving the accuracy of network capacity prediction from the source of data.

[0053] In addition, the software test configuration parameters may include power configuration, frequency configuration, bandwidth configuration, switching parameters, load balancing parameters, etc. It should be noted that the software test configuration parameters may be used to reflect network operation conditions and optimize performance.

[0054] Specifically, each fusion evaluation parameter can be used as a node to construct the original relationship graph of the rule. The original relationship graph is input into the GCN network, and any parameter is selected from the performance parameters on the network management side. The connection edge calculation is performed on this parameter and the remaining fusion evaluation parameters to obtain the corresponding association weight. It should be noted that for a graph structure data set G, there are N nodes, each of which has its own characteristics. The characteristics of these nodes can be set to form a matrix x of size N*D, where D represents the hidden state dimension of each node. In addition, the relationship between each node can also be extracted as an N*N relationship matrix x, also called an adjacency matrix, and y represents the validation set. A is used to represent the information of the edge between nodes, and the feature weight w is calculated according to the minimum value of the loss function minE((x,y),y′)∼p(ℓ(y′,y^(x,y))). Figure 3 is a comparative schematic diagram of feature reconstruction by integrating evaluation parameters provided by the present invention, such as Figure 3As shown in the figure, in the graph before feature reconstruction, the connecting edges between the fusion evaluation parameters (x1, x2, ...) are the same, which cannot convey the relationship between nodes, that is, it cannot express the intrinsic connection between user trajectories and spatial features. In the graph after feature reconstruction, each node is no longer a regular Euclidean shape, but more of a relationship between nodes.

[0055] Figure 4 is a schematic diagram of a feature reconstruction relationship map obtained based on the maximum number of RRC users provided by the present invention, such as Figure 4 As shown, taking the maximum number of RRC users as an example, the association weights between the maximum number of RRC users and the remaining fusion evaluation parameters are calculated respectively. After feature reconstruction, the weights between the maximum number of RRC users and each feature can be obtained. A hierarchical structure is formed after weighted operation. The first layer has the largest weight, such as the number of users and the number of historical users, traffic, and power configuration. The second layer has the second largest weight, such as the number of users and downlink PRB utilization, uplink PRB utilization, hanging height, switching parameters, and so on. In detail, Table 1 shows the association weights between any network management side performance parameter and each fusion evaluation parameter except any network management side performance parameter, as shown in Table 1 below: Table 1

[0056] It can be understood that by calculating each feature separately, relationship graphs of different features can be obtained respectively.

[0057] In view of the fact that traditional prediction algorithms are insensitive to short-term business and cannot provide timely warnings, the method provided in the embodiments of the present invention puts part of the computing power offline for completion, greatly reducing the computing cost during the guarantee period, and outputting the capacity prediction results in a timely manner.

[0058] Based on any of the above embodiments, Figure 5 FIG. 3 is a flow chart of the network capacity prediction method provided by the present invention. Figure 5 As shown, the method includes: First, input data, including network management side performance data, base station side physical information, and software test configuration information, to build the underlying feature data set. Then, the underlying feature data set is reconstructed for spatiotemporal features. The spatiotemporal feature weight W is output based on the GCN feature reconstruction module.

[0059] Further, we enter the short-term business prediction stage. Specifically, in the training stage of the GCN prediction model, we first initialize the training set, then train the GCN model based on the training set, generate a prediction function, and perform time series prediction and user trajectory prediction based on the prediction function. It should be noted that in order to judge the prediction effect of the model, the mean absolute error (MAE) is used, as shown in the following formula:

[0060] In the formula, is the actual flow value of the i-th test sample, is the predicted flow value of the i-th test sample; n is the total number of test samples, error MAE≤5% is a reasonable value range, if MAE>5%, the model deviation correction module is used for auxiliary correction. In addition, the training data input terminal x is the input data of the previous step, and the output terminal Test y is the output data of the previous step, and the training set: validation set = 7:3.

[0061] In the application stage of the GCN prediction model, the network capacity prediction value can be obtained based on the GCN prediction model by inputting the network management side performance data, base station side physical information, software measurement configuration information and spatiotemporal feature weights. It should be noted that the GCN prediction model has the following characteristics, including: (1) Time series feature prediction: Based on the feature enhancement of the GCN proxy model, the mapping relationship between basic environmental variables and performance indicators is established, and the proxy model of PRB utilization and traffic, RRC maximum number of users, etc. is output, which can be achieved by calling the short-term business prediction model function. (2) User trajectory prediction: Based on the data of the first 10 nodes, the data of the last 5 nodes is predicted. The existing historical data predicts the increase or decrease trend of the next time node compared with the previous moment for daily regional sites with an accuracy of more than 90%. It can be presented geographically through the prediction function combined with the longitude and latitude information of the cell. Figure 6 is a schematic diagram of the relationship between the user flow trajectory and the physical characteristics provided by the present invention, such as Figure 6 As shown: the direction indicated by the arrow is the user flow trajectory, which corresponds to from base station A at time t1 to base station B at time t2, and finally to base station C at time t3.

[0062] Based on any of the above embodiments, Figure 7 is a schematic diagram of the structure of the network capacity prediction device provided by the present invention, such as Figure 7 As shown, the device comprises: An acquisition unit 710 acquires fusion evaluation parameters of the area to be predicted; The prediction unit 720 inputs the fusion evaluation parameter and the association weight between any network management side performance parameter in the fusion evaluation parameter and each fusion evaluation parameter other than the any network management side performance parameter into a capacity prediction model, so as to obtain a network capacity prediction value of the area to be predicted outputted by the capacity prediction model; The capacity prediction model is obtained by training an initial capacity prediction model based on sample fusion evaluation parameters.

[0063] The device provided by the embodiment of the present invention inputs the association weights between the fusion evaluation parameter and any network management side performance parameter and each fusion evaluation parameter excluding any network management side performance parameter into the capacity prediction model, and indicates the capacity prediction model to capture the connection between the user trajectory change and the spatial change in the fusion evaluation parameter through the connection weight. This realizes fast and accurate network capacity prediction.

[0064] Figure 8 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 8 As shown, the electronic device may include: a processor 810, a communication interface 820, a memory 830 and a communication bus 840, wherein the processor 810, the communication interface 820 and the memory 830 communicate with each other through the communication bus 840. The processor 810 may call the logic instructions in the memory 830 to execute the network capacity prediction method, which includes: obtaining the fusion evaluation parameters of the area to be predicted; inputting the fusion evaluation parameters and the association weights between any network management side performance parameters in the fusion evaluation parameters and each fusion evaluation parameter other than any network management side performance parameters into the capacity prediction model, and obtaining the network capacity prediction value of the area to be predicted output by the capacity prediction model; the capacity prediction model is obtained by training the initial capacity prediction model based on the sample fusion evaluation parameters.

[0065] In addition, the logic instructions in the above-mentioned memory 830 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0066] On the other hand, the present invention also provides a computer program product, which includes a computer program, and the computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the network capacity prediction method provided by the above-mentioned methods, and the method includes: obtaining a fusion evaluation parameter of the area to be predicted; inputting the fusion evaluation parameter and the association weight between any network management side performance parameter in the fusion evaluation parameter and each fusion evaluation parameter other than any network management side performance parameter into a capacity prediction model, so as to obtain a network capacity prediction value of the area to be predicted output by the capacity prediction model; the capacity prediction model is obtained by training an initial capacity prediction model based on the sample fusion evaluation parameter.

[0067] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the network capacity prediction method provided by the above-mentioned methods, the method comprising: obtaining fusion evaluation parameters of the area to be predicted; inputting the fusion evaluation parameters and the association weights between any network management side performance parameters in the fusion evaluation parameters and each fusion evaluation parameter other than any network management side performance parameter into a capacity prediction model, so as to obtain a network capacity prediction value of the area to be predicted output by the capacity prediction model; the capacity prediction model is obtained by training an initial capacity prediction model based on sample fusion evaluation parameters.

[0068] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.

[0069] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0070] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. 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. However, 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 embodiments of the present invention.

Claims

1. A network capacity prediction method, characterized in that: include: Obtain fusion evaluation parameters of the area to be predicted; Input the fusion evaluation parameter and the association weight between any network management side performance parameter in the fusion evaluation parameter and each fusion evaluation parameter other than the any network management side performance parameter into a capacity prediction model, so as to obtain a network capacity prediction value of the area to be predicted outputted by the capacity prediction model; The capacity prediction model is obtained by training an initial capacity prediction model based on sample fusion evaluation parameters.

2. The network capacity prediction method according to claim 1, characterized in that: The capacity prediction model includes a graph convolution unit and a gated recurrent unit; The step of inputting the fusion evaluation parameter and the association weight between any network management side performance parameter in the fusion evaluation parameter and each fusion evaluation parameter other than the any network management side performance parameter into a capacity prediction model, and obtaining a network capacity prediction value of the area to be predicted outputted by the capacity prediction model, comprises: Extracting parameter features of the fusion evaluation parameters at the historical moment based on the graph convolution unit and the associated weights; Performing feature processing on the parameter features of the historical moments based on the gated recurrent unit and the self-attention mechanism to obtain the parameter features of the current moment; Based on the parameter characteristics at the current moment, the network capacity prediction value is obtained.

3. The network capacity prediction method according to claim 2, characterized in that: The gated recurrent unit includes an update gate, a reset gate, a candidate hidden gate, and an update hidden gate; The feature processing of the parameter features of the historical moments based on the gated recurrent unit and the self-attention mechanism to obtain the parameter features of the current moment includes: Based on the update gate and the first attention weight corresponding to the update gate, feature retaining the parameter feature of the historical moment to obtain a first retained feature; Based on the reset gate and the second attention weight corresponding to the reset gate, feature discarding is performed on the parameter feature of the historical moment to obtain a second retained feature; Based on the candidate hidden door and the third attention weight corresponding to the candidate hidden door, feature processing is performed on the parameter feature of the historical moment and the second retained feature to obtain a third retained feature; Based on the updated hidden gate, feature processing is performed on the first retained feature, the second retained feature, and the third retained feature to obtain the current moment parameter feature.

4. The network capacity prediction method according to claim 2, characterized in that: The obtaining the network capacity prediction value based on the parameter characteristics at the current moment includes: Based on the prediction function and the parameter characteristics at the current moment, an initial capacity prediction value is obtained; Performing network capacity prediction on the area to be predicted based on the network adjustment parameters of the area to be predicted to obtain a capacity adjustment prediction value; The initial capacity prediction value is corrected based on the capacity adjustment prediction value to obtain the network capacity prediction value.

5. The network capacity prediction method according to claim 4, characterized in that: The network adjustment parameters include at least the number of system switching times between base stations and individual cell offset parameters.

6. The network capacity prediction method according to any one of claims 1 to 5, characterized in that: The fusion evaluation parameters also include base station side physical parameters and software test configuration parameters; The step of acquiring the association weight between any network management side performance parameter in the fusion evaluation parameter and each fusion evaluation parameter other than the any network management side performance parameter comprises: Based on the graph convolutional network, connection edges are calculated for any of the network management side performance parameters and each of the fusion evaluation parameters, so as to obtain an association weight between any of the network management side performance parameters and each of the fusion evaluation parameters except any of the network management side performance parameters.

7. A network capacity prediction device, characterized in that: include: An acquisition unit, which acquires fusion evaluation parameters of the area to be predicted; The prediction unit inputs the fusion evaluation parameter and the association weight between any network management side performance parameter in the fusion evaluation parameter and each fusion evaluation parameter other than the any network management side performance parameter into a capacity prediction model, so as to obtain a network capacity prediction value of the area to be predicted outputted by the capacity prediction model; The capacity prediction model is obtained by training an initial capacity prediction model based on sample fusion evaluation parameters.

8. An electronic 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 network capacity prediction method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the network capacity prediction method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the network capacity prediction method according to any one of claims 1 to 6 is implemented.

Citation Information

Patent Citations

  • Network element capacity analysis and prediction method based on deep neural networks

    CN108471353A

  • 5G network quality automatic evaluation method based on machine learning

    CN114648053A

  • Network capacity prediction method and device, electronic equipment and storage medium

    CN115473821A

  • Method and device for generating communication network index parameter prediction model

    CN115879029A

  • Data physical fusion driven natural gas dehydration device space-time diagram convolutional network parameter prediction method

    CN116090341A

Cited By

  • Model parameter fusion method and device

    CN119720101A