Traffic prediction method and device, and computer readable storage medium

Abstract

This invention provides a traffic prediction method, apparatus, and computer-readable storage medium, relating to the field of network communication technology, and is used to solve the problem of inaccurate traffic prediction in existing technologies. The method includes: acquiring a first Radio Resource Control (RRC) connection count, a first Physical Resource Block (PRB) utilization rate, and first traffic data of a base station within a first historical time period; inputting the first RRC connection count and the first PRB utilization rate into a pre-constructed first traffic prediction model to obtain a first predicted traffic of the base station within a preset future time period; inputting the first traffic data into a pre-constructed second traffic prediction model to obtain a second predicted traffic of the base station within the preset future time period; and inputting the first predicted traffic and the second predicted traffic into a pre-constructed data fusion model to obtain a traffic prediction result of the base station within the preset future time period. This invention can obtain more accurate traffic prediction results than existing technologies.

Description

Technical Field

[0001] This invention relates to the field of network communication technology, and in particular to a traffic prediction method, apparatus, and computer-readable storage medium. Background Technology

[0002] Currently, network traffic forecasting by operators, whether based on experience or using a single model, can lead to inaccurate forecasts. Using a single model fails to fully exploit the value of traffic forecasting data and does not integrate the impact of historical base station data on future traffic predictions from multiple perspectives, thus failing to provide accurate and effective forecast results. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to address the above-mentioned shortcomings of the prior art by providing a traffic prediction method, apparatus and computer-readable storage medium, so as to solve the problem that the prior art may lead to inaccurate traffic prediction due to its reliance on experience or the use of a single model.

[0004] In a first aspect, the present invention provides a traffic prediction method, the method comprising:

[0005] Acquire the number of first Radio Resource Control (RRC) connections, the utilization rate of the first Physical Resource Block (PRB), and the first traffic data of the base station within the first historical time period;

[0006] The first RRC connection number and the first PRB utilization rate are input into the pre-built first traffic prediction model to obtain the first predicted traffic of the base station in a preset future time period.

[0007] The first traffic data is input into a pre-built second traffic prediction model to obtain the second predicted traffic of the base station within a preset future time period.

[0008] The first predicted traffic and the second predicted traffic are input into a pre-built data fusion model to obtain the traffic prediction results of the base station within a preset future time period.

[0009] Optionally, the first traffic prediction model is a linear model, and the second traffic prediction model is a machine learning model.

[0010] Optionally, the step of inputting the first traffic data into a pre-built second traffic prediction model to obtain the second predicted traffic of the base station within a preset future time period specifically includes:

[0011] Multiple first traffic data of different lengths are extracted according to different elasticity coefficients. The first traffic data of each length are combined into prediction input parameters of a preset step size according to the corresponding elasticity coefficient, and then input into the pre-built second traffic prediction model to obtain multiple different second predicted traffic of the base station within a preset future time period.

[0012] Optionally, the method further includes:

[0013] Acquire the number of second radio resource control (RRC) connections, the utilization rate of the second physical resource block (PRB), and the second traffic data of the base station within the second historical time period;

[0014] The second RRC connection count, the second PRB utilization rate, and the second traffic data are used to construct the first traffic prediction model, the second traffic prediction model, and the data fusion model.

[0015] The first historical time period is determined based on the preset future time period and the division of the second historical time period when constructing the first traffic prediction model, the second traffic prediction model and the data fusion model.

[0016] Optionally, before inputting the first RRC connection number and the first PRB utilization rate into a pre-constructed first traffic prediction model to obtain the first predicted traffic of the base station within a preset future time period, the method further includes:

[0017] Divide the second historical time period into multiple previous and subsequent historical time periods, construct a linear multiple regression model between the second RRC connection number and second PRB utilization rate of the previous historical time period and the second traffic data of the subsequent historical time period, and use the linear multiple regression model as the first traffic prediction model.

[0018] Optionally, before inputting the first traffic data into a pre-built second traffic prediction model to obtain the second predicted traffic of the base station within a preset future time period, the method further includes:

[0019] From each previous historical time period, multiple second flow data of different lengths are extracted according to different elasticity coefficients. The second flow data of each length are used to form training input parameters with a preset step size according to the corresponding elasticity coefficient. The second flow data of the corresponding next historical time period is used as the training result parameters. A GRU (Gated Recurrent Unit) gated recurrent neural network model is trained using the training input parameters and the training result parameters. The GRU gated recurrent neural network model is used as the second flow prediction model.

[0020] Optionally, before inputting the first predicted traffic and the second predicted traffic into a pre-built data fusion model to obtain the traffic prediction results of the base station within a preset future time period, the method further includes:

[0021] The second RRC connection number and the second PRB utilization rate of each previous historical time period are input into the linear multiple regression model to obtain the first simulated predicted flow.

[0022] The second flow data of each step length in each previous historical time period is input into the GRU gated recurrent neural network model to obtain multiple second simulated predicted flows.

[0023] The first simulated predicted traffic and the plurality of second simulated predicted traffic are input into the Lasso lasso algorithm, and the fusion coefficient of the Lasso lasso algorithm is obtained by fitting it with the second traffic data of the corresponding historical time period. The Lasso lasso algorithm with the fusion coefficient is used as the data fusion model.

[0024] In a second aspect, the present invention provides a flow prediction device, comprising:

[0025] The acquisition module is used to acquire the first radio resource control (RRC) connection number, the first physical resource block (PRB) utilization rate, and the first traffic data of the base station within the first historical time period.

[0026] The first prediction module, connected to the acquisition module, is used to input the first RRC connection number and the first PRB utilization rate into a pre-built first traffic prediction model to obtain the first predicted traffic of the base station within a preset future time period.

[0027] The second prediction module, connected to the acquisition module, is used to input the first traffic data into a pre-built second traffic prediction model to obtain the second predicted traffic of the base station within a preset future time period.

[0028] The fusion module, connected to the first prediction module and the second prediction module, is used to input the first predicted traffic and the second predicted traffic into a pre-built data fusion model to obtain the traffic prediction results of the base station within a preset future time period.

[0029] Thirdly, the present invention provides a traffic prediction device, including a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the traffic prediction method as described above.

[0030] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when run by a processor, implements the traffic prediction method as described above.

[0031] This invention provides a traffic prediction method, apparatus, and computer-readable storage medium. By inputting the historical RRC connection count and PRB utilization of a base station into a first traffic prediction model and inputting the historical traffic data of the base station into a second traffic prediction model, the influence of the RRC connection count and PRB utilization on traffic prediction is explored. The historical traffic data is used to predict future traffic data. Finally, by fusing the two models to predict traffic, a more accurate traffic prediction result than that of a single model is obtained. Attached Figure Description

[0032] Figure 1 This is a flowchart of a traffic prediction method according to an embodiment of the present invention;

[0033] Figure 2 This is a schematic diagram of the structure of a flow prediction device according to an embodiment of the present invention;

[0034] Figure 3 This is a schematic diagram of another flow prediction device according to an embodiment of the present invention. Detailed Implementation

[0035] To enable those skilled in the art to better understand the technical solution of the present invention, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0036] It is understood that the specific embodiments and accompanying drawings described herein are merely for explaining the invention and are not intended to limit the invention.

[0037] It is understood that, without conflict, the various embodiments and features in the embodiments of the present invention can be combined with each other.

[0038] It is understood that, for ease of description, only the parts related to the present invention are shown in the accompanying drawings, while the parts unrelated to the present invention are not shown in the drawings.

[0039] It is understood that each unit or module involved in the embodiments of the present invention may correspond to only one entity structure, or may be composed of multiple entity structures, or multiple units or modules may be integrated into one entity structure.

[0040] It is understood that, without conflict, the functions and steps marked in the flowcharts and block diagrams of this invention may occur in a different order than that marked in the accompanying drawings.

[0041] It is understood that the flowcharts and block diagrams of this invention illustrate the possible architecture, functions, and operations of systems, apparatuses, devices, and methods according to various embodiments of the invention. Each block in the flowchart or block diagram may represent a unit, module, program segment, or code, containing executable instructions for implementing the specified function. Furthermore, each block or combination of blocks in the block diagram and flowchart can be implemented using a hardware-based system to achieve the specified function, or using a combination of hardware and computer instructions.

[0042] It is understood that the units and modules involved in the embodiments of the present invention can be implemented by software or by hardware. For example, the units and modules can be located in a processor.

[0043] Example 1:

[0044] like Figure 1 As shown, the present invention provides a traffic prediction method, the method comprising:

[0045] S11. Obtain the first radio resource control (RRC) connection number, the first physical resource block (PRB) utilization rate, and the first traffic data of the base station within the first historical time period.

[0046] S12. Input the first RRC connection number and the first PRB utilization rate into the pre-built first traffic prediction model to obtain the first predicted traffic of the base station in a preset future time period.

[0047] S13. Input the first traffic data into the pre-built second traffic prediction model to obtain the second predicted traffic of the base station in a preset future time period;

[0048] S14. Input the first predicted traffic and the second predicted traffic into a pre-built data fusion model to obtain the traffic prediction results of the base station within a preset future time period.

[0049] Specifically, in this embodiment, the pre-built traffic prediction model includes a first traffic prediction model, a second traffic prediction model, and a data fusion model, which form a T-shaped structure. Historical data for prediction is input into the first traffic prediction model and the second traffic prediction model respectively. Finally, the results of the two models are simultaneously input into the data fusion model to obtain the final traffic prediction result. When applied to the actual prediction of base station traffic, the historical RRC connection count and PRB utilization rate related to base station and network traffic are input into the first traffic prediction model, and the historical traffic data of the base station is input into the second traffic prediction model. The impact of RRC connection count and PRB utilization rate on traffic prediction is explored, and future traffic data is predicted using historical traffic data. Finally, by fusing the two models to predict traffic, a more accurate traffic prediction result than that of a single model is obtained.

[0050] Optionally, the first traffic prediction model is a linear model, and the second traffic prediction model is a machine learning model.

[0051] Specifically, in this embodiment, to address the issue that shallow linear model methods cannot learn relatively accurate traffic data and deep machine learning models suffer from overlearning in order to obtain accurate traffic prediction results, a T-shaped traffic prediction model is established by combining the advantages of both. Based on the linear model and the machine learning model, a data fusion model is used to improve the accuracy of the linear model's prediction results and reduce the learning accuracy requirements of the machine learning model, thus achieving the effect of both in-depth learning and avoiding overlearning.

[0052] Optionally, the step of inputting the first traffic data into a pre-built second traffic prediction model to obtain the second predicted traffic of the base station within a preset future time period specifically includes:

[0053] Multiple first traffic data of different lengths are extracted according to different elasticity coefficients. The first traffic data of each length are combined into prediction input parameters of a preset step size according to the corresponding elasticity coefficient, and then input into the pre-built second traffic prediction model to obtain multiple different second predicted traffic of the base station within a preset future time period.

[0054] Specifically, in this embodiment, it is assumed that the preset step size of the second traffic prediction model is S, and the elasticity coefficient is 0-N. The elasticity coefficient indicates how many data points are empty. For example, the first traffic data is the historical traffic of the past nine days: [a1, a2, a3, a4, a5, a6, a7, a8, a9], and the second predicted traffic is the traffic of the next three days [X1, X2, X3]. When the elasticity coefficient is 0, the prediction input parameters are [a1, a2, a3, a4, a5, a6, a7, a8, a9], and when the elasticity coefficient is 1, the prediction input parameters are [nan, a1, a2]. [a1, a2, a3, a4, a5, a6, a7, a8], and so on, we can obtain N prediction input parameters with different actual data lengths. These are then input into the second traffic prediction model to obtain N different predicted values ​​X11, X12, ..., X1N for X1. At this time, the first traffic prediction model will also obtain a predicted value X1L for X1. By fusing these N+1 data through the data fusion model and using different fusion coefficients, the data with strong correlation to the true value of the prediction result is filtered out to obtain a more accurate prediction result for X1. The same method can be used to predict X2 and X3 as for X1.

[0055] Optionally, the method further includes:

[0056] Acquire the number of second radio resource control (RRC) connections, the utilization rate of the second physical resource block (PRB), and the second traffic data of the base station within the second historical time period;

[0057] The second RRC connection count, the second PRB utilization rate, and the second traffic data are used to construct the first traffic prediction model, the second traffic prediction model, and the data fusion model.

[0058] The first historical time period is determined based on the preset future time period and the division of the second historical time period when constructing the first traffic prediction model, the second traffic prediction model and the data fusion model.

[0059] Specifically, in this embodiment, a T-shaped traffic prediction model needs to be constructed in advance using traffic-related data from a second historical time period. The second historical time period is usually longer than the first historical time period. By collecting long-term traffic-related data from base stations, the correlation between historical data and future traffic data is obtained, and a model for predicting future traffic data is constructed. When constructing the model, the prediction requirements are considered. For different prediction requirements for future time periods, such as predicting hourly traffic changes in the next day, total traffic in the next day, or average daily traffic in the next month, the second historical time period is divided differently when constructing the traffic prediction model. When predicting traffic, the first historical time period is determined based on the time length divided when constructing the model.

[0060] After obtaining the first Radio Resource Control (RRC) connection count, first Physical Resource Block (PRB) utilization rate, and first traffic data of the base station within the first historical time period, or after obtaining the second RRC connection count, second PRB utilization rate, and second traffic data of the base station within the second historical time period, the method further includes:

[0061] The acquired RRC connection count, PRB utilization, and traffic data are preprocessed, including anomaly detection, smoothing, and normalization.

[0062] Specifically, in this embodiment, in order to improve the accuracy of the model, data from the second historical time period will be obtained from multiple base stations to build the model. The original data obtained from different base stations may have different data lengths, different time lengths, and different patterns between holiday data and normal data. Therefore, it is necessary to preprocess and unify the format before it can be used to build the model. When predicting the data of the first historical time period of the base stations, the data also needs to be processed in the same way.

[0063] Optionally, before inputting the first RRC connection number and the first PRB utilization rate into a pre-constructed first traffic prediction model to obtain the first predicted traffic of the base station within a preset future time period, the method further includes:

[0064] Divide the second historical time period into multiple previous and subsequent historical time periods, construct a linear multiple regression model between the second RRC connection number and second PRB utilization rate of the previous historical time period and the second traffic data of the subsequent historical time period, and use the linear multiple regression model as the first traffic prediction model.

[0065] Specifically, in this embodiment, the first traffic prediction model adopts a linear multiple regression model to explore the relationship between the number of base station RRC connections, PRB utilization and traffic. The linear multiple regression model can be constructed by collecting multiple sets of related data and then performing curve fitting.

[0066] Optionally, before inputting the first traffic data into a pre-built second traffic prediction model to obtain the second predicted traffic of the base station within a preset future time period, the method further includes:

[0067] From each previous historical time period, multiple second flow data of different lengths are extracted according to different elasticity coefficients. The second flow data of each length are used to form training input parameters with a preset step size according to the corresponding elasticity coefficient. The second flow data of the corresponding next historical time period is used as the training result parameters. A GRU gated recurrent neural network model is obtained by training with the training input parameters and the training result parameters. The GRU gated recurrent neural network model is used as the second flow prediction model.

[0068] Specifically, in this embodiment, the second flow prediction method corresponding to the aforementioned elastic step size also adopts the method of preset step size and elastic coefficient when constructing the second flow prediction model. The second flow prediction model is specifically a GRU gated recurrent neural network model. Its construction process adopts the training method. Once the model parameters that meet the preset convergence conditions are obtained, the training stops. Subsequently, the GRU gated recurrent neural network with corresponding model parameters is used for second flow prediction. By using the elastic step size, more of the correlation between flow over time can be explored, such as the different demand for flow on weekdays and holidays.

[0069] Optionally, before inputting the first predicted traffic and the second predicted traffic into a pre-built data fusion model to obtain the traffic prediction results of the base station within a preset future time period, the method further includes:

[0070] The second RRC connection number and the second PRB utilization rate of each previous historical time period are input into the linear multiple regression model to obtain the first simulated predicted flow.

[0071] The second flow data of each step length in each previous historical time period is input into the GRU gated recurrent neural network model to obtain multiple second simulated predicted flows.

[0072] The first simulated predicted traffic and the plurality of second simulated predicted traffic are input into the Lasso lasso algorithm, and the fusion coefficient of the Lasso lasso algorithm is obtained by fitting it with the second traffic data of the corresponding historical time period. The Lasso lasso algorithm with the fusion coefficient is used as the data fusion model.

[0073] Specifically, in this embodiment, the data fusion model is the Lasso lasso algorithm. To obtain the fusion coefficient of the Lasso lasso algorithm, the corresponding input data and result data are obtained according to the division of the second historical time period by the constructed model. That is, the data of each historical time period before the second historical time period is used as input data to obtain simulated predicted flow, and the flow data of the historical time period after the second historical time period is used as result data to obtain the fusion coefficient. The first simulated predicted flow is obtained by using the fitted linear multiple regression model, and multiple second simulated predicted flows are obtained by using the trained GRU gated recurrent neural network model. Finally, the simulated predicted flow and the result data are fitted to obtain the fusion coefficient. The characteristic of the Lasso lasso algorithm is that it selects variables with strong features into the model and discards some variables with weak features, which greatly improves the accuracy of the prediction results.

[0074] The traffic prediction results obtained through the steps described in Example 1 can be used to provide operators with accurate and effective decision-making basis, make decisions on possible network traffic warnings, capacity expansion, and other situations, rationally allocate resources, and detect abnormal traffic warnings in real time.

[0075] Example 2:

[0076] like Figure 2 As shown, Embodiment 2 of the present invention provides a flow prediction device, comprising:

[0077] The acquisition module 11 is used to acquire the first radio resource control (RRC) connection number, the first physical resource block (PRB) utilization rate, and the first traffic data of the base station within the first historical time period.

[0078] The first prediction module 12 is connected to the acquisition module 11 and is used to input the first RRC connection number and the first PRB utilization rate into the pre-built first traffic prediction model to obtain the first predicted traffic of the base station in a preset future time period.

[0079] The second prediction module 13 is connected to the acquisition module 11 and is used to input the first traffic data into a pre-built second traffic prediction model to obtain the second predicted traffic of the base station within a preset future time period.

[0080] The fusion module 14, connected to the first prediction module 12 and the second prediction module 13, is used to input the first predicted traffic and the second predicted traffic into a pre-built data fusion model to obtain the traffic prediction results of the base station within a preset future time period.

[0081] Optionally, the first traffic prediction model is a linear model, and the second traffic prediction model is a machine learning model.

[0082] Optionally, the second prediction module 13 is specifically used for:

[0083] Multiple first traffic data of different lengths are extracted according to different elasticity coefficients. The first traffic data of each length are combined into prediction input parameters of a preset step size according to the corresponding elasticity coefficient, and then input into the pre-built second traffic prediction model to obtain multiple different second predicted traffic of the base station within a preset future time period.

[0084] Optionally, the acquisition module 11 is further configured to:

[0085] Acquire the number of second radio resource control (RRC) connections, the utilization rate of the second physical resource block (PRB), and the second traffic data of the base station within the second historical time period;

[0086] The second RRC connection count, the second PRB utilization rate, and the second traffic data are used to construct the first traffic prediction model, the second traffic prediction model, and the data fusion model.

[0087] The first historical time period is determined based on the preset future time period and the division of the second historical time period when constructing the first traffic prediction model, the second traffic prediction model and the data fusion model.

[0088] Optionally, the acquisition module 11 specifically includes:

[0089] The acquisition unit is used to acquire the number of RRC connections, PRB utilization and traffic data of the base station within the first / second historical time period;

[0090] A preprocessing unit, connected to the acquisition unit, is used to preprocess the acquired RRC connection count, PRB utilization, and traffic data. The preprocessing includes anomaly detection, smoothing, and normalization.

[0091] Optionally, the device further includes:

[0092] The first construction module is used to divide the second historical time period into multiple previous historical time periods and subsequent historical time periods, construct a linear multiple regression model between the second RRC connection number and the second PRB utilization rate of the previous historical time period and the second traffic data of the subsequent historical time period, and use the linear multiple regression model as the first traffic prediction model.

[0093] Optionally, the device further includes:

[0094] The second construction module is used to extract multiple second flow data of different lengths from each previous historical time period according to different elasticity coefficients, to form training input parameters of a preset step size for each length of second flow data according to the corresponding elasticity coefficient, and to use the second flow data of the corresponding next historical time period as training result parameters. The GRU gated recurrent neural network model is trained using the training input parameters and the training result parameters, and the GRU gated recurrent neural network model is used as the second flow prediction model.

[0095] Optionally, the device further includes a third building module, specifically comprising:

[0096] The first simulation unit is used to input the second RRC connection number and the second PRB utilization rate of each previous historical time period into the linear multiple regression model to obtain the first simulated predicted flow.

[0097] The second simulation unit is used to input the second flow data of each step length in each previous historical time period into the GRU gated recurrent neural network model to obtain multiple second simulated predicted flows.

[0098] The fitting unit, connected to the first simulation unit and the second simulation unit, is used to input the first simulated predicted flow and the plurality of second simulated predicted flows into the Lasso lasso algorithm, fit it with the second flow data of the corresponding historical subsequent time period to obtain the fusion coefficient of the Lasso lasso algorithm, and use the Lasso lasso algorithm with the fusion coefficient as the data fusion model.

[0099] Example 2 provides a traffic prediction device, which corresponds to the traffic prediction method in Example 1. It can fully mine historical data of the base station, accurately predict the future traffic of the base station by using the historical RRC connection number, PRB utilization rate and historical traffic of the base station, and then use the predicted traffic to provide accurate and effective basis for operator decision-making.

[0100] Example 3:

[0101] like Figure 3As shown, Embodiment 3 of the present invention provides a traffic prediction device, including a memory 10 and a processor 20. The memory 10 stores a computer program. When the processor 20 runs the computer program stored in the memory 10, the processor 20 executes the traffic prediction method as described in Embodiment 1.

[0102] The memory 10 is connected to the processor 20. The memory 10 can be a flash memory, a read-only memory, or another type of memory. The processor 20 can be a central processing unit or a microcontroller.

[0103] Example 4:

[0104] Embodiment 4 of the present invention provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the traffic prediction method as described in Embodiment 1.

[0105] The computer-readable storage medium includes volatile or non-volatile, removable or non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, computer program modules, or other data). Computer-readable storage media include, but are not limited to, RAM (Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory or other memory technologies, CD-ROM (Compact Disc Read-Only Memory), DVD or other optical disc storage, cartridges, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer.

[0106] Embodiments 1-4 of the present invention provide a traffic prediction method, apparatus, and computer-readable storage medium. By inputting the historical RRC connection count and PRB utilization of the base station into a first traffic prediction model and inputting the historical traffic data of the base station into a second traffic prediction model, the influence of the RRC connection count and PRB utilization on traffic prediction is explored, and the historical traffic data is used to predict future traffic data. Finally, by fusing the two models to predict traffic, a more accurate traffic prediction result than that of a single model is obtained.

[0107] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A flow prediction method, characterized in that, The method includes: Obtain the first Radio Resource Control (RRC) connection count, the first Physical Resource Block (PRB) utilization rate, and the first traffic data of the base station within the first historical time period. Obtain the number of second radio resource control (RRC) connections, the utilization rate of the second physical resource block (PRB), and the second traffic data of the base station within the second historical time period. The second RRC connection count, the second PRB utilization rate, and the second traffic data are used to construct the first traffic prediction model, the second traffic prediction model, and the data fusion model. The first historical time period is determined based on the preset future time period and the division of the second historical time period when constructing the first traffic prediction model, the second traffic prediction model and the data fusion model; The first RRC connection number and the first PRB utilization are input into the pre-built first traffic prediction model to obtain the first predicted traffic of the base station in a preset future time period. The first traffic prediction model is a linear multiple regression model. The first traffic data is input into a pre-built second traffic prediction model to obtain the second predicted traffic of the base station within a preset future time period. The second traffic prediction model is a GRU-gated recurrent neural network model, specifically including: Multiple first traffic data of different lengths are extracted according to different elasticity coefficients. The elasticity coefficient indicates how many data are empty. The first traffic data of each length are combined into prediction input parameters of a preset step size according to the corresponding elasticity coefficient, and then input into the pre-built second traffic prediction model to obtain multiple different second predicted traffic of the base station within a preset future time period. The second historical period is divided into multiple preceding and succeeding historical periods. The second RRC connection number and the second PRB utilization rate of each previous historical time period are input into the linear multiple regression model to obtain the first simulated predicted flow. The second flow data of each step length in each previous historical time period is input into the GRU gated recurrent neural network model to obtain multiple second simulated predicted flows. The first simulated predicted flow and the plurality of second simulated predicted flows are input into the Lasso lasso algorithm, and the fusion coefficient of the Lasso lasso algorithm is obtained by fitting it with the second flow data of the corresponding historical time period. The Lasso lasso algorithm with the fusion coefficient is used as the data fusion model. The first predicted traffic and multiple second predicted traffic are input into the pre-constructed data fusion model. The fusion coefficient is used to filter the data in the first predicted traffic and multiple second predicted traffic that are strongly correlated with the actual value of the prediction result, and discard some data with weak correlation, so as to obtain the traffic prediction result of the base station within a preset future time period.

2. The method according to claim 1, characterized in that, Before inputting the first RRC connection number and the first PRB utilization rate into the pre-constructed first traffic prediction model to obtain the first predicted traffic of the base station within a preset future time period, the method further includes: A linear multiple regression model is constructed between the second RRC connection number and the second PRB utilization rate in the previous historical time period and the second traffic data in the next historical time period. The linear multiple regression model is used as the first traffic prediction model.

3. The method according to claim 2, characterized in that, Before inputting the first traffic data into a pre-constructed second traffic prediction model to obtain the second predicted traffic of the base station within a preset future time period, the method further includes: From each previous historical time period, multiple second flow data of different lengths are extracted according to different elasticity coefficients. The second flow data of each length are used to form training input parameters with a preset step size according to the corresponding elasticity coefficient. The second flow data of the corresponding next historical time period is used as the training result parameters. A GRU gated recurrent neural network model is obtained by training with the training input parameters and the training result parameters. The GRU gated recurrent neural network model is used as the second flow prediction model.

4. A flow prediction device, characterized in that, include: The acquisition module is used to acquire the first Radio Resource Control (RRC) connection count, the first Physical Resource Block (PRB) utilization rate, and the first traffic data of the base station within the first historical time period. Obtain the number of second radio resource control (RRC) connections, the utilization rate of the second physical resource block (PRB), and the second traffic data of the base station within the second historical time period. The second RRC connection count, the second PRB utilization rate, and the second traffic data are used to construct the first traffic prediction model, the second traffic prediction model, and the data fusion model. The first historical time period is determined based on the preset future time period and the division of the second historical time period when constructing the first traffic prediction model, the second traffic prediction model and the data fusion model; The first prediction module, connected to the acquisition module, is used to input the first RRC connection number and the first PRB utilization rate into a pre-constructed first traffic prediction model to obtain the first predicted traffic of the base station within a preset future time period. The first traffic prediction model is a linear multiple regression model. The second prediction module, connected to the acquisition module, is used to input the first traffic data into a pre-constructed second traffic prediction model to obtain the second predicted traffic of the base station within a preset future time period. The second prediction model is a GRU-gated recurrent neural network model, specifically used for: Multiple first traffic data of different lengths are extracted according to different elasticity coefficients. The elasticity coefficient indicates how many data are empty. The first traffic data of each length are combined into prediction input parameters of a preset step size according to the corresponding elasticity coefficient, and then input into the pre-built second traffic prediction model to obtain multiple different second predicted traffic of the base station within a preset future time period. The module is used to divide the second historical time period into multiple preceding and succeeding historical time periods. The second RRC connection number and the second PRB utilization rate of each previous historical time period are input into the linear multiple regression model to obtain the first simulated predicted flow. The second flow data of each step length in each previous historical time period is input into the GRU gated recurrent neural network model to obtain multiple second simulated predicted flows. The first simulated predicted flow and the plurality of second simulated predicted flows are input into the Lasso lasso algorithm, and the fusion coefficient of the Lasso lasso algorithm is obtained by fitting it with the second flow data of the corresponding historical time period. The Lasso lasso algorithm with the fusion coefficient is used as the data fusion model. The fusion module, connected to the first prediction module and the second prediction module, is used to input the first predicted traffic and multiple second predicted traffic into the pre-built data fusion model, and to filter the data in the first predicted traffic and multiple second predicted traffic that are strongly correlated with the actual value of the prediction result and discard some data with weak correlation with the feature through the fusion coefficient, so as to obtain the traffic prediction result of the base station within a preset future time period.

5. A flow prediction device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor performs the traffic prediction method as described in any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the traffic prediction method as described in any one of claims 1-3.

Images