Network index prediction method and device

Abstract

The invention provides a network index prediction method and a network index prediction device, and belongs to the technical field of communication. The network index prediction method comprises: obtaining a quality index of a cellular network cell and a first parameter related to the quality index, the first parameter and the quality index having a causal relationship, and the change of the first parameter affecting the quality index; obtaining first historical data of the quality index, second historical data of the first parameter and prediction data of the first parameter; and inputting the first historical data, the second historical data and the prediction data of the first parameter into a network index prediction model corresponding to the quality index to obtain a first prediction result of the quality index. According to the technical scheme, the network index prediction precision can be improved.

Description

Technical Field

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

[0002] Cellular networks provide convenient network access services for user devices such as mobile phones. Such services are built on a large number of physical devices (including base stations). Cellular network service providers carefully select the location of base stations to improve the quality of service (QoS) while making full use of the resources of each base station. The channel resources in the cellular network are limited, and a large number of users in the cell will lead to an increase in congestion and a decrease in service quality. At this time, the parameters of the cellular network base station can be adjusted and optimized to improve the service quality, but the existing cellular network base stations rely on the operation and maintenance engineers to select new parameters based on their personal experience, and the selected parameters may not be appropriate. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a network indicator prediction method and device, which can improve the network indicator prediction accuracy, thereby assisting in adjusting and optimizing the configuration parameters of cellular network base stations.

[0004] To solve the above technical problems, the embodiments of the present invention provide the following technical solutions:

[0005] On the one hand, an embodiment of the present invention provides a network indicator prediction method, comprising:

[0006] Acquire a quality indicator of a cellular network cell and a first parameter related to the quality indicator, where the first parameter has a causal relationship with the quality indicator, and a change in the first parameter affects the quality indicator;

[0007] Acquire first historical data of the quality indicator, second historical data of the first parameter, and predicted data of the first parameter;

[0008] The first historical data, the second historical data, and the predicted data of the first parameter are input into a network indicator prediction model corresponding to the quality indicator to obtain a first prediction result of the quality indicator.

[0009] In some embodiments, when the quality indicator changes with the change of the configuration parameters of the cellular network cell, after obtaining the first prediction result of the quality indicator, the method further includes:

[0010] Obtaining an adjustment amount of a configuration parameter of the cellular network cell;

[0011] Determining a correction coefficient corresponding to the quality indicator according to the adjustment amount;

[0012] The first prediction result is corrected using the correction coefficient to obtain a second prediction result of the quality indicator.

[0013] In some embodiments, the quality indicator is a cell load, the configuration parameters include a transmission power and a cell offset of the cellular network cell, and determining a correction coefficient corresponding to the quality indicator according to the adjustment amount includes:

[0014] According to the transmission power adjustment amount δ and the cell offset adjustment amount δ o Determine the first coefficient

[0015] The correction coefficient is determined according to the first coefficient, the distribution angle interval of multiple adjacent cells of the cellular network cell, and the center angle corresponding to the boundary arc of the service area of ​​the cellular network cell.

[0016] In some embodiments, the correction coefficient α(β|θ) is determined according to the following formula:

[0017]

[0018] Among them, γ is the central angle corresponding to the arc of the service area boundary of the cellular network cell, and θ is the distribution angle interval of multiple adjacent cells of the cellular network cell.

[0019] In some embodiments, the acquiring the quality indicator of the cellular network cell and the first parameter related to the quality indicator comprises:

[0020] Establishing a directed acyclic graph according to the causal relationship between the plurality of quality indicators and the plurality of first parameters;

[0021] The parent node of the quality indicator in the directed acyclic graph is used as a first parameter related to the quality indicator.

[0022] In some embodiments, the method further comprises the step of training the network indicator prediction model, and training the network indicator prediction model comprises:

[0023] Acquire the third historical data of the first time period and the fourth historical data of the second time period of the quality indicator, acquire the fifth historical data of the first time period and the sixth historical data of the second time period of the first parameter, wherein the start time of the second time period is after the end time of the first time period;

[0024] The third historical data, the fifth historical data, and the sixth historical data are used as inputs of the network indicator prediction model, and the fourth historical data is used as output of the network indicator prediction model to train the network indicator prediction model.

[0025] In some embodiments, inputting the first historical data, the second historical data, and the predicted data of the first parameter into a network indicator prediction model corresponding to the quality indicator to obtain a first prediction result of the quality indicator includes:

[0026] Inputting the first historical data and the second historical data into an encoder of the network indicator prediction model;

[0027] The output of the encoder, the average value of the first historical data, the second historical data and the predicted data are input into the decoder of the network indicator prediction model to obtain the first prediction result.

[0028] The embodiment of the present invention also provides a network indicator prediction device, including:

[0029] A first acquisition module, configured to acquire a quality indicator of a cellular network cell and a first parameter related to the quality indicator, wherein the first parameter has a causal relationship with the quality indicator, and a change in the first parameter affects the quality indicator;

[0030] A second acquisition module, used to acquire first historical data of the quality indicator, second historical data of the first parameter and predicted data of the first parameter;

[0031] A processing module is used to input the first historical data, the second historical data and the predicted data of the first parameter into a network indicator prediction model corresponding to the quality indicator to obtain a first prediction result of the quality indicator.

[0032] In some embodiments, when the quality indicator changes with changes in configuration parameters of the cellular network cell, the apparatus further comprises:

[0033] A correction module is used to obtain the adjustment amount of the configuration parameters of the cellular network cell; determine the correction coefficient corresponding to the quality indicator according to the adjustment amount; and use the correction coefficient to correct the first prediction result to obtain a second prediction result of the quality indicator.

[0034] In some embodiments, the quality indicator is a cell load, and the configuration parameters include a transmission power and a cell offset of the cellular network cell.

[0035] The correction module is specifically configured to adjust the transmission power according to the adjustment amount δ and the cell offset adjustment amount δ. o Determine the first coefficient The correction coefficient is determined according to the first coefficient, the distribution angle interval of multiple adjacent cells of the cellular network cell, and the center angle corresponding to the boundary arc of the service area of ​​the cellular network cell.

[0036] In some embodiments, the correction module is specifically used to determine the correction coefficient α(β|θ) according to the following formula:

[0037]

[0038] Among them, γ is the central angle corresponding to the arc of the service area boundary of the cellular network cell, and θ is the distribution angle interval of multiple adjacent cells of the cellular network cell.

[0039] In some embodiments, the first acquisition module is specifically used to establish a directed acyclic graph according to the causal relationship between multiple quality indicators and multiple first parameters; and use the parent node of the quality indicator in the directed acyclic graph as the first parameter related to the quality indicator.

[0040] In some embodiments, the device further comprises a training module.

[0041] The training module is specifically used to obtain the third historical data of the first time period and the fourth historical data of the second time period of the quality indicator, obtain the fifth historical data of the first time period and the sixth historical data of the second time period of the first parameter, and the start time of the second time period is after the end time of the first time period; use the third historical data, the fifth historical data, and the sixth historical data as the input of the network indicator prediction model, and use the fourth historical data as the output of the network indicator prediction model to train the network indicator prediction model.

[0042] In some embodiments, the processing module is specifically used to input the first historical data and the second historical data into the encoder of the network indicator prediction model; input the output of the encoder, the average value of the first historical data, the second historical data and the predicted data into the decoder of the network indicator prediction model to obtain the first prediction result.

[0043] An embodiment of the present invention also provides a network indicator prediction device, including a memory, a processor, and a computer program stored in the memory and executable on the processor; when the processor executes the program, the network indicator prediction method as described above is implemented.

[0044] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the network indicator prediction method described above are implemented.

[0045] The embodiments of the present invention have the following beneficial effects:

[0046] In the above scheme, the quality indicator is predicted based on the historical data of the quality indicator, the historical data of the first parameter related to the quality indicator and the predicted data, which can improve the prediction accuracy of the quality indicator. The operation and maintenance engineers can adjust the configuration parameters of the cellular network base station according to the prediction results of the quality indicator to ensure the service quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A schematic diagram of a flow chart of a network indicator prediction method according to an embodiment of the present invention;

[0048] Figure 2 A cause-and-effect relationship diagram in a cellular network according to an embodiment of the present invention;

[0049] Figure 3 This is a schematic diagram of time series prediction based on the Transformer model according to an embodiment of the present invention;

[0050] Figure 4 It is a schematic diagram of the service area of ​​an isotropic radiator antenna;

[0051] Figure 5 A schematic diagram of prediction results after configuration parameters of a cellular network cell are adjusted;

[0052] Figure 6 A schematic diagram of the structure of a network indicator prediction device according to an embodiment of the present invention;

[0053] Figure 7 The figure is a schematic diagram of the composition of a network indicator prediction device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to make the technical problems, technical solutions and advantages to be solved by the embodiments of the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0055] Existing cellular network base stations rely on operation and maintenance engineers to select new parameters based on their personal experience, and the selected parameters may not be appropriate. In order to predict the effect of parameter adjustment, the existing solutions include the following two:

[0056] The first solution is to actually apply the new parameters on the base station, and then check whether the base station meets expectations after collecting new data. The advantage of this solution is that the results of parameter adjustment can be accurately understood. However, this solution takes a long time to obtain the parameter effect, and there is a risk of applying unreasonable parameters to the actual base station, reducing user experience.

[0057] The second solution is to use a data-driven approach to learn the relationship between the results of parameter adjustment, the amount of parameter change, and the cell status before parameter adjustment. The advantage of this solution is that it can quickly predict the results of parameter adjustment. However, for a large number of monitoring indicators and variable parameters, this solution has a large demand for historical data and requires sufficient historical data, otherwise it is difficult to ensure accurate prediction results.

[0058] The embodiments of the present invention provide a network indicator prediction method and device, which can improve the network indicator prediction accuracy, thereby assisting in adjusting and optimizing the parameters of cellular network base stations.

[0059] The embodiment of the present invention provides a network indicator prediction method, such as Figure 1 As shown, including:

[0060] Step 101: Acquire a quality indicator of a cellular network cell and a first parameter related to the quality indicator, wherein the first parameter has a causal relationship with the quality indicator, and a change in the first parameter affects the quality indicator;

[0061] Step 102: Acquire first historical data of the quality indicator, second historical data of the first parameter, and predicted data of the first parameter;

[0062] Step 103: input the first historical data, the second historical data and the predicted data of the first parameter into a network indicator prediction model corresponding to the quality indicator to obtain a first prediction result of the quality indicator.

[0063] In this embodiment, the quality indicator is predicted based on the historical data of the quality indicator, the historical data of the first parameter related to the quality indicator, and the predicted data, so as to improve the prediction accuracy of the quality indicator. The operation and maintenance engineers can adjust the configuration parameters of the cellular network base station according to the prediction results of the quality indicator to ensure the service quality.

[0064] When the quality indicator is not affected by changes in configuration parameters of the cellular network cell, the prediction result of the quality indicator can be directly obtained by using the network indicator prediction model.

[0065] In some embodiments, the acquiring the quality indicator of the cellular network cell and the first parameter related to the quality indicator comprises:

[0066] Establishing a directed acyclic graph according to the causal relationship between the plurality of quality indicators and the plurality of first parameters;

[0067] The parent node of the quality indicator in the directed acyclic graph is used as a first parameter related to the quality indicator.

[0068] The monitoring indicators of a cell in a cellular network (i.e., the above-mentioned quality indicators) can be divided into several categories according to their meanings, such as counting the number of user devices accessing the cell and the load of their behaviors, and counting the interference of electromagnetic wave signal strength that is not related to the signal. Each category of monitoring indicators is called a monitoring variable, and the monitoring variables can be organized into a causal relationship graph G=<V,E> . Where V is the set of all monitoring variables, E is the set of causal relationships between monitoring variables, and G is a directed acyclic graph. In a specific example, when the monitoring indicators are load W, interference I, and service quality Q, the established directed acyclic graph is as follows Figure 2 As shown. The parent variable (i.e., the first parameter) set of the monitoring variable X in the causal relationship graph is Pa(X), then Pa(X) = {Y|(Y→X)∈E}, and the parent variable of the monitoring variable X is the parent node of the monitoring variable X in the causal relationship graph. Figure 2 For example, load W and interference I are parent variables of service quality Q. In addition to the existing monitoring indicators, this embodiment can also add timestamp information as a separate variable T to the causal relationship graph G, and the causal relationship with it as the starting node reflects the time periodicity of the corresponding monitoring variable.

[0069] In this embodiment, based on the causal relationship graph formed by the monitoring variables, the monitoring variables in the causal relationship graph may be traversed in topological order, and the prediction results of each monitoring variable may be accumulated.

[0070] For nodes without parent variables in the causal relationship graph, the expected value of the time period to be predicted is used as the prediction result. For example, for the time variable T, the corresponding time series can be generated using the timestamp of the time period to be predicted.

[0071] For the node with a parent variable in the causal relationship diagram, that is, the monitoring variable X, obtain the historical data x before the period to be predicted (t-L) ,…,x (t-1) , where t represents the start time of the period to be predicted, L represents the size of the time window used for reference, and x (t-1) That is, the last sampled data before the start time of the period to be predicted. Further, the historical data pa of the parent variable Pa(X) of the monitoring variable X before the period to be predicted is obtained. (t-L) (X),…,pa (t-1) (X), and the predicted data of the parent variable Pa(X) Wherein, O represents the length of the period to be predicted. Specifically, the network indicator prediction model corresponding to the parent variable Pa(X) can be used to obtain the prediction data of the parent variable Pa(X). When the monitoring variable X corresponds to multiple parent variables Pa(X), the prediction data of multiple parent variables Pa(X) can be obtained in sequence according to the traversal order (topological order) of the causal relationship graph.

[0072] The historical data x(t-L) ,…,x (t-1) 、Historical data (t-L) (X),…,pa (t-1) (X), prediction data Input the network indicator prediction model corresponding to the monitoring variable X to obtain the value of the monitoring variable X in the future As the first prediction result of the monitoring variable X.

[0073] Specifically, inputting the first historical data, the second historical data, and the predicted data of the first parameter into a network indicator prediction model corresponding to the quality indicator to obtain a first prediction result of the quality indicator includes:

[0074] Inputting the first historical data and the second historical data into an encoder of the network indicator prediction model;

[0075] The output of the encoder, the average value of the first historical data, the second historical data and the predicted data are input into the decoder of the network indicator prediction model to obtain the first prediction result.

[0076] like Figure 3 As shown, the historical data x (t-L) ,…,x (t-1) 、Historical data (t-L) (X),…,pa (t-1) (X) Input the encoder, and convert the encoder output and historical data x (t-L) ,…,x (t-1) Average value, historical data x (t-L / 2) ,…,x (t-1) 、Historical data (t-L / 2) (X),…,pa (t-1) (X), predicted data pa (t) (X),…,pa (t+O-1) (X) inputting the decoder of the network indicator prediction model to obtain a first prediction result

[0077] In this embodiment, the network indicator prediction model corresponding to the monitoring variable X can be obtained by training with historical data in advance, and the training of the network indicator prediction model includes:

[0078] Acquire the third historical data of the first time period and the fourth historical data of the second time period of the quality indicator, acquire the fifth historical data of the first time period and the sixth historical data of the second time period of the first parameter, wherein the start time of the second time period is after the end time of the first time period;

[0079] The third historical data, the fifth historical data, and the sixth historical data are used as inputs of the network indicator prediction model, and the fourth historical data is used as output of the network indicator prediction model to train the network indicator prediction model.

[0080] In this embodiment, the network indicator prediction model can be trained using historical data of multiple time periods. That is, the network indicator prediction model can be trained using historical data of multiple groups of time periods, each group of time periods includes a first time period and a second time period corresponding to the first time period, such as in one group of time periods, 2023-06-20 00:00:00 to 2023-06-22 00:00:00 is used as the first time period, and 2023-06-22 00:15:00 to 2023-06-24 00:15:00 is used as the second time period.

[0081] In this embodiment, the network indicator prediction models corresponding to different monitoring variables X are trained independently, and the values ​​of Pa(X) in the period to be predicted in the actual data are used without using the prediction results of other prediction models, which can improve the accuracy of the prediction results. The network indicator prediction model of this embodiment can select a regression model or a deep learning model, and the supervision signal can select the prediction error, or it can be combined with the supervision signals of different tasks such as the reconstruction error. During most of the time when the cellular network base station is running, the cell parameters remain unchanged. This embodiment can use the data of the period without parameter adjustment to predict the monitoring variable X, which can make the prediction result more stable.

[0082] In some embodiments, when the quality indicator changes with the change of the configuration parameters of the cellular network cell, after obtaining the first prediction result of the quality indicator, the method further includes:

[0083] Obtaining an adjustment amount of a configuration parameter of the cellular network cell;

[0084] Determining a correction coefficient corresponding to the quality indicator according to the adjustment amount;

[0085] The first prediction result is corrected using the correction coefficient to obtain a second prediction result of the quality indicator.

[0086] like Figure 2 When the load W shown is used as a monitoring indicator, it changes with the configuration parameters of the cellular network cell (including transmission power and cell offset). Therefore, after obtaining the first prediction result of the quality indicator using the network indicator prediction model, the first prediction result needs to be corrected to improve the accuracy of the prediction result of the quality indicator.

[0087] For the monitoring variable X affected by the configuration parameters of the cellular network cell, obtain the historical data x before the period to be predicted(t-L) ,…,x (t-1) , where t represents the start time of the period to be predicted, L represents the size of the time window used for reference, and x (t-1) That is, the last sampled data before the start time of the period to be predicted. Further, the historical data pa of the parent variable Pa(X) of the monitoring variable X before the period to be predicted is obtained. (t-L) (X),…,pa (t-1) (X), and the predicted data of the parent variable Pa(X) Wherein, O represents the length of the period to be predicted. Specifically, the prediction data of the parent variable Pa(X) can be obtained by using the network indicator prediction model corresponding to the parent variable Pa(X). When the monitoring variable X corresponds to multiple parent variables Pa(X), the prediction data of multiple parent variables Pa(X) can be obtained according to the traversal order (topological order) of the causal relationship graph.

[0088] The historical data x (t-L) ,…,x (t-1) 、Historical data (t-L) (X),…,pa (t-1) (X), prediction data Input the network indicator prediction model corresponding to the monitoring variable X to obtain the value of the monitoring variable X in the future As the first prediction result of the monitoring variable X. The first prediction result is corrected based on the correction coefficient obtained by parameter adjustment to obtain the second prediction result

[0089] The adjustment of configuration parameters has different effects on different monitoring variables. For example, increasing the transmission power of a cell will not increase the strength of the interference signal. Therefore, in this embodiment, the prediction result of the monitoring variable X is corrected only when the monitoring variable X is affected by the adjustment of the configuration parameters.

[0090] Table 1 shows the classification of monitoring indicators of cells in cellular networks.

[0091] Table 1

[0092]

[0093] In this embodiment, the Transformer model may be used as the network indicator prediction model. Figure 3 It is a Transformer-based time series prediction model architecture. The encoder processes historical data and generates corresponding latent representations, and the decoder concatenates data from the following two time periods: recent historical data; the initial estimate of the variable to be predicted (such as the historical average) or the value of the parent variable in the causal relationship graph (such as the time variable in the time period to be predicted).

[0094] The decoder also uses the encoder's output as input to calculate mutual attention and generate another latent representation. Finally, the final latent representation is restored to a time series through the projection module, replacing the initial estimate of the variable to be predicted with the new predicted value. The decoder's output has the same length as the input, and only the period to be predicted needs to be intercepted as the output of the model.

[0095] The following takes the quality indicator as the cell load, and the configuration parameters of the cellular network base station including the transmission power and the cell offset (Cell Individual Offset, CIO) of the cellular network cell as an example to further introduce the technical solution of this embodiment.

[0096] Transmission power and cell offset are two commonly used optimization parameters. Transmission power is a basic parameter of a cell. The higher the transmission power of a cell, the higher the reference signal received power (RSRP) measured by the user equipment from the cell signal, and the user equipment tends to access the cell with the highest RSRP. CIO can be used to control the switching of user equipment between cells to achieve better load balancing. In related technologies, when a user equipment observes a target cell t with a higher RSRP than the source cell s, it will switch to the target cell t, that is, after satisfying formula (1), it will switch from the source cell to the target cell, where RSRP t RSRP is the reference signal received power of the target cell t measured by the user equipment s CIO is the reference signal received power of the source cell s measured by the user equipment. t→s is the CIO from the target cell t to the source cell s, CIO s→t is the CIO from the source cell s to the target cell t, and Hys is a hysteresis parameter used to reduce the number of handovers of cell edge user equipment. Therefore, CIO and Hys can affect when a user equipment is to be handed over from one cell to another.

[0097] RSRP t +CIO t→s >RSRP s +CIO s→t +Hys (1)

[0098] Cell-edge user equipment is more sensitive to the adjustment of transmission power and CIO. For a cell-edge user equipment U and the cell A it accesses, after reducing the transmission power or CIO of A, the RSRP measured by the user equipment U may satisfy formula (1), thereby triggering the user equipment U to switch to the adjacent cell B and disconnect from A. In this process, the load of A is reduced by the usage of U, while the load of B is increased. If the owner of the user equipment U is also using another user equipment at the same time, the load changes in the above process will be different. Since the two user equipments are in the same position, the measured RSRP will also be the same. If the same parameter change causes the two user equipments to switch at the same time, the load of A will be reduced by 2 times the usage of U. To this end, this embodiment introduces the concept of usage density and models the cell load as the integral of the usage of user equipment in the cell service area, that is, W = ∫∫ S ρ(x,y)dx dy. Where S is the area of ​​the cell A, and ρ(x,y) is the usage density function. Since it is impossible to infer the value of the usage density function at various locations based on the existing data, this embodiment assumes that the average usage density at the cell edge is the same as that of the entire cell, which is Under this assumption, this embodiment infers the change of cell load by calculating the change of cell area S′ / S.

[0099] The hysteresis parameter in formula (1) expands the boundary between cells into a strip with a width. In order to simplify the analysis process, a virtual boundary that satisfies the following conditions can be defined: the number of user devices accessing other cells on the side close to the antenna (i.e., the antenna of the base station) within the virtual boundary is the same as the number of user devices accessing the cell on the other side of the virtual boundary. Based on this virtual boundary, the impact of the hysteresis parameter on monitoring indicators such as the number of user devices that measure cell load can be ignored.

[0100] Formula (2) defines a virtual boundary. Different cells have the same RSRP offset by CIO for user equipment on the boundary. It should be noted that the parameters of adjacent cells (such as CIO t→s ) will also participate in determining the cell boundary, but this embodiment analyzes a single cell as the granularity and does not consider the data of adjacent cells. For this reason, it is assumed that the adjacent cells and the given cell have the same initial parameters.

[0101] RSRP t +CIO t→s =RSRP s +CIO s→t (2)

[0102] In the operation and maintenance of cellular networks, transmission power and CIO are measured in dBm. Accordingly, when the transmission power changes from τdBm to τ+δdBm and the CIO changes from o dBm to o+δ odBm, the RSRP measured by a user equipment after CIO offset will be R becomes P′ R =βP R ,in

[0103] Next, the antennas of the two base stations are analyzed as two isotropic radiators, such as Figure 4 The antenna of base station A and the antenna of base station B in the figure. When the same initial parameters are configured, β = 1, the service areas of A and B on the plane are divided by the perpendicular bisector of the line connecting the two ( Figure 4 The vertical line in the figure). When the RSRP of A decreases after CIO offset, β<1. When the parameter change of A satisfies β<1, the service area of ​​A shrinks and the area is reduced. Let P be the initial transmission power of A. Since the power density of electromagnetic waves is inversely proportional to the square of the distance to the point energy source, for a user device on the virtual boundary of A and B defined by formula (2), its coordinates (a, b) satisfy formula (3).

[0104]

[0105] Where 2r is the distance between base station A and base station B.

[0106] Formula (4) can be obtained from formula (3), which shows that the virtual boundary when β<1 is Figure 4 China-Israel is the center of the circle, A circle with a radius of Figure 4 The dotted circle on the left side of the center.

[0107]

[0108] Further consider that A has multiple adjacent cells distributed around A with the same angle interval θ<π (for example, θ=60°). A's service area is further reduced to several Figure 4 The shaded area in the figure. Let γ be the central angle of the arc corresponding to the boundary of the shaded area. From the geometric relationship, we can see that γ satisfies formula ((5). Based on γ, formula ((6) calculates the area of ​​the shaded area.

[0109]

[0110]

[0111] The value of θ can be obtained by the deployment of base stations, and the value of β can be obtained by the parameter changes of transmission power and CIO; however, the model with a single cell as the analysis granularity cannot obtain r required by formula (6). In order to eliminate the influence of r, this embodiment calculates the ratio of the cell area after and before the parameter adjustment, that is, α(β|θ)=Sshaded / [r 2 tan(θ / 2)]. When β>1, this embodiment simplifies the calculation to α(β|θ)=2-α(1 / β|θ). Formula (7) presents the complete calculation formula of α(β|θ).

[0112]

[0113] The antennas in cellular networks are generally directional. Therefore, the plane boundary between two adjacent cells of the same base station can be regarded as the ray emitted from the base station, so that S can be directly estimated using α(β|θ) ′ / S, and use α(β|θ) as a multiplier to act on the predicted cell load value without parameter adjustment, that is, That is, use α(β|θ) as the correction coefficient to predict the cell load value Correction, cell load forecast value The first prediction result of the cell load is obtained by using the network indicator prediction model. In this embodiment, it is only necessary to pre-train the Transformer model using the data of the historical period without parameter adjustment, and use the Transformer model to obtain the first prediction result of the cell load, and then use the correction coefficient to adjust the cell load prediction value. The correction does not require data from the parameter adjustment period, which can improve the accuracy of network indicator prediction.

[0114] Figure 5 This is the prediction result of the maximum number of RRC connections for a certain cell in this embodiment. Figure 5 The dashed straight line is used as the boundary in the figure. The left side shows the maximum number of RRC connections in a certain cell when the transmission power is 18.2 dBm, and the right side shows the maximum number of RRC connections in a certain cell after the transmission power is adjusted to 12.2 dBm. Figure 5 The solid line in the middle is the actual data, and the dotted line is the predicted data after the transmission power is adjusted. The predicted data includes the predicted data Y1 obtained using GT (Graphical Transformer) (i.e., the first prediction result obtained using the technical solution of this embodiment), the predicted data Y2 obtained using GBRT (Gradient Boosting Regression Tree), and the corrected predicted data Y3 (i.e., the second prediction result obtained using the technical solution of this embodiment). It can be seen that the corrected predicted data Y3 is closest to the actual data.

[0115] The embodiment of the present invention also provides a network indicator prediction device, such as Figure 6 As shown, including:

[0116] A first acquisition module 21, configured to acquire a quality indicator of a cellular network cell and a first parameter related to the quality indicator, wherein the first parameter has a causal relationship with the quality indicator, and a change in the first parameter affects the quality indicator;

[0117] A second acquisition module 22, used to acquire first historical data of the quality indicator, second historical data of the first parameter and predicted data of the first parameter;

[0118] The processing module 23 is used to input the first historical data, the second historical data and the predicted data of the first parameter into the network indicator prediction model corresponding to the quality indicator to obtain a first prediction result of the quality indicator.

[0119] In this embodiment, the quality indicator is predicted based on the historical data of the quality indicator, the historical data of the first parameter related to the quality indicator, and the predicted data, so as to improve the prediction accuracy of the quality indicator. The operation and maintenance engineers can adjust the configuration parameters of the cellular network base station according to the prediction results of the quality indicator to ensure the service quality.

[0120] In some embodiments, when the quality indicator changes with changes in configuration parameters of the cellular network cell, the apparatus further comprises:

[0121] A correction module is used to obtain the adjustment amount of the configuration parameters of the cellular network cell; determine the correction coefficient corresponding to the quality indicator according to the adjustment amount; and use the correction coefficient to correct the first prediction result to obtain a second prediction result of the quality indicator.

[0122] In some embodiments, the quality indicator is a cell load, and the configuration parameters include a transmission power and a cell offset of the cellular network cell.

[0123] The correction module is specifically configured to adjust the transmission power according to the adjustment amount δ and the cell offset adjustment amount δ. o Determine the first coefficient The correction coefficient is determined according to the first coefficient, the distribution angle interval of multiple adjacent cells of the cellular network cell, and the center angle corresponding to the boundary arc of the service area of ​​the cellular network cell.

[0124] In some embodiments, the correction module is specifically used to determine the correction coefficient α(β|θ) according to the following formula:

[0125]

[0126] Among them, γ is the central angle corresponding to the arc of the service area boundary of the cellular network cell, and θ is the distribution angle interval of multiple adjacent cells of the cellular network cell.

[0127] In some embodiments, the first acquisition module 21 is specifically used to establish a directed acyclic graph according to the causal relationship between multiple quality indicators and multiple first parameters; and use the parent node of the quality indicator in the directed acyclic graph as the first parameter related to the quality indicator.

[0128] In some embodiments, the device further comprises a training module.

[0129] The training module is specifically used to obtain the third historical data of the first time period and the fourth historical data of the second time period of the quality indicator, obtain the fifth historical data of the first time period and the sixth historical data of the second time period of the first parameter, and the start time of the second time period is after the end time of the first time period; use the third historical data, the fifth historical data, and the sixth historical data as the input of the network indicator prediction model, and use the fourth historical data as the output of the network indicator prediction model to train the network indicator prediction model.

[0130] In some embodiments, the processing module 23 is specifically used to input the first historical data and the second historical data into the encoder of the network indicator prediction model; input the output of the encoder, the average value of the first historical data, the second historical data and the predicted data into the decoder of the network indicator prediction model to obtain the first prediction result.

[0131] The embodiment of the present invention also provides a network indicator prediction device, such as Figure 7 As shown, it includes a memory 31, a processor 32, and a computer program stored in the memory 31 and executable on the processor 32; when the processor 32 executes the program, the network indicator prediction method as described above is implemented.

[0132] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the network indicator prediction method described above are implemented.

[0133] Computer readable media include permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage terminal devices to be detected or any other non-transmission media that can be used to store information that can be accessed by the computer terminal device to be detected. As defined in this article, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carriers.

[0134] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A network indicator prediction method, characterized in that: include: Acquire a quality indicator of a cellular network cell and a first parameter related to the quality indicator, where the first parameter has a causal relationship with the quality indicator, and a change in the first parameter affects the quality indicator; Acquire first historical data of the quality indicator, second historical data of the first parameter, and predicted data of the first parameter; The first historical data, the second historical data, and the predicted data of the first parameter are input into a network indicator prediction model corresponding to the quality indicator to obtain a first prediction result of the quality indicator.

2. The network indicator prediction method according to claim 1, characterized in that: When the quality indicator changes with the configuration parameter of the cellular network cell, after obtaining a first prediction result of the quality indicator, the method further includes: Obtaining an adjustment amount of a configuration parameter of the cellular network cell; Determining a correction coefficient corresponding to the quality indicator according to the adjustment amount; The first prediction result is corrected using the correction coefficient to obtain a second prediction result of the quality indicator.

3. The network indicator prediction method according to claim 2, characterized in that: The quality indicator is a cell load, the configuration parameters include a transmission power and a cell offset of the cellular network cell, and determining a correction coefficient corresponding to the quality indicator according to the adjustment amount includes: According to the transmission power adjustment amount δ and the cell offset adjustment amount δ o Determine the first coefficient The correction coefficient is determined according to the first coefficient, the distribution angle interval of multiple adjacent cells of the cellular network cell, and the center angle corresponding to the boundary arc of the service area of ​​the cellular network cell.

4. The network indicator prediction method according to claim 3, characterized in that: The correction coefficient α(β|θ) is determined according to the following formula: Among them, γ is the central angle corresponding to the arc of the service area boundary of the cellular network cell, and θ is the distribution angle interval of multiple adjacent cells of the cellular network cell.

5. The network indicator prediction method according to claim 1, characterized in that: The acquiring of the quality indicator of the cellular network cell and the first parameter related to the quality indicator comprises: Establishing a directed acyclic graph according to the causal relationship between the plurality of quality indicators and the plurality of first parameters; The parent node of the quality indicator in the directed acyclic graph is used as a first parameter related to the quality indicator.

6. The network indicator prediction method according to claim 1, characterized in that: The method further comprises the step of training the network indicator prediction model, wherein the training of the network indicator prediction model comprises: Acquire the third historical data of the first time period and the fourth historical data of the second time period of the quality indicator, acquire the fifth historical data of the first time period and the sixth historical data of the second time period of the first parameter, wherein the start time of the second time period is after the end time of the first time period; The third historical data, the fifth historical data, and the sixth historical data are used as inputs of the network indicator prediction model, and the fourth historical data is used as output of the network indicator prediction model to train the network indicator prediction model.

7. The network indicator prediction method according to claim 1, characterized in that: The step of inputting the first historical data, the second historical data, and the predicted data of the first parameter into a network indicator prediction model corresponding to the quality indicator to obtain a first prediction result of the quality indicator includes: Inputting the first historical data and the second historical data into an encoder of the network indicator prediction model; The output of the encoder, the average value of the first historical data, the second historical data and the predicted data are input into the decoder of the network indicator prediction model to obtain the first prediction result.

8. A network indicator prediction device, characterized in that: include: A first acquisition module, configured to acquire a quality indicator of a cellular network cell and a first parameter related to the quality indicator, wherein the first parameter has a causal relationship with the quality indicator, and a change in the first parameter affects the quality indicator; A second acquisition module, used to acquire first historical data of the quality indicator, second historical data of the first parameter and predicted data of the first parameter; A processing module is used to input the first historical data, the second historical data and the predicted data of the first parameter into a network indicator prediction model corresponding to the quality indicator to obtain a first prediction result of the quality indicator.

9. The network indicator prediction device according to claim 8, characterized in that: When the quality indicator changes with the change of the configuration parameters of the cellular network cell, the apparatus further includes: A correction module is used to obtain the adjustment amount of the configuration parameters of the cellular network cell; determine the correction coefficient corresponding to the quality indicator according to the adjustment amount; and use the correction coefficient to correct the first prediction result to obtain a second prediction result of the quality indicator.

10. The network indicator prediction device according to claim 9, characterized in that: The quality indicator is the cell load, and the configuration parameters include the transmission power and cell offset of the cellular network cell. The correction module is specifically configured to adjust the transmission power according to the adjustment amount δ and the cell offset adjustment amount δ. o Determine the first coefficient The correction coefficient is determined according to the first coefficient, the distribution angle interval of multiple adjacent cells of the cellular network cell, and the center angle corresponding to the boundary arc of the service area of ​​the cellular network cell.

11. The network indicator prediction device according to claim 10, characterized in that: The correction module is specifically used to determine the correction coefficient α(β|θ) according to the following formula: Among them, γ is the central angle corresponding to the arc of the service area boundary of the cellular network cell, and θ is the distribution angle interval of multiple adjacent cells of the cellular network cell.

12. The network indicator prediction device according to claim 8, characterized in that: The first acquisition module is specifically used to establish a directed acyclic graph according to the causal relationship between multiple quality indicators and multiple first parameters; and use the parent node of the quality indicator in the directed acyclic graph as the first parameter related to the quality indicator.

13. The network indicator prediction device according to claim 8, characterized in that: The device also includes a training module, The training module is specifically used to obtain the third historical data of the first time period and the fourth historical data of the second time period of the quality indicator, obtain the fifth historical data of the first time period and the sixth historical data of the second time period of the first parameter, and the start time of the second time period is after the end time of the first time period; use the third historical data, the fifth historical data, and the sixth historical data as the input of the network indicator prediction model, and use the fourth historical data as the output of the network indicator prediction model to train the network indicator prediction model.

14. The network indicator prediction device according to claim 8, characterized in that: The processing module is specifically used to input the first historical data and the second historical data into the encoder of the network indicator prediction model; input the output of the encoder, the average value of the first historical data, the second historical data and the predicted data of the first parameter into the decoder of the network indicator prediction model to obtain the first prediction result.

15. A network indicator prediction device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor; characterized in that: When the processor executes the program, the network indicator prediction method according to any one of claims 1 to 7 is implemented.

16. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the network indicator prediction method as described in any one of claims 1 to 7 are implemented.

Images