Method, apparatus and computer program product for load pre-evaluation of a wireless cell

Abstract

The application discloses a kind of load pre-evaluation method, device and computer program product of wireless cell.Therein, the method includes: obtaining the target measurement report and target operating parameter data of the wireless cell to be evaluated;From the target measurement report, determine the dynamic communication performance characteristics of the wireless cell to be evaluated;From the target operating parameter data, determine the static communication performance characteristics of the wireless cell to be evaluated;Using cell load pre-evaluation model to analyze dynamic communication performance characteristics and static communication performance characteristics, obtain the target load index of the wireless cell to be evaluated, wherein the cell load pre-evaluation model is pre-trained by machine learning using multiple sets of training data, each set of training data in the multiple sets of training data includes: the dynamic communication performance characteristics, static communication performance characteristics and pre-calibrated sample load index of the same sample wireless cell.The application solves the technical problem of low accuracy of wireless cell load pre-evaluation.

Description

Technical Field

[0001] This invention relates to the field of wireless communication, and more specifically, to a method, apparatus, and computer program product for pre-assessing the load of a wireless cell. Background Technology

[0002] In wireless networks, the uneven load distribution among wireless cells within the same coverage area is a prominent issue due to factors such as cell power, frequency band, azimuth angle, downtilt angle, bandwidth, terminal capabilities, user distribution, and service distribution. Therefore, balancing the load among wireless cells within the area is of paramount importance.

[0003] Currently, load balancing technology is commonly used to achieve load balancing between wireless cells. However, due to the large number of parameters affecting load balancing, the actual balancing effect often does not match expectations. Therefore, it is very important to accurately predict the load of wireless cells after parameter adjustment.

[0004] Industry research on wireless cell load pre-assessment assesses wireless cell load based on the number of users in the wireless cell. For example:

[0005] This patent's method assumes a linear relationship between cell load and the number of users. It first calculates the number of users after load balancing and then divides it by the cell's maximum user capacity to pre-assess the cell load. However, this method fails to consider the non-linear relationship between cell load and user count, making it difficult to accurately predict cell load based solely on user count and thus compromising the accuracy of the pre-assessment.

[0006] Assuming a non-linear relationship between cell load and the number of users, this patent's method first calculates the number of users after load balancing, and then uses a third-order polynomial fitting equation to pre-evaluate the cell load. However, this method, while modeling the non-linear relationship between cell load and users using a third-order polynomial fitting equation, still fails to accurately model this relationship. Therefore, it's difficult to accurately predict cell load based solely on user count, and the accuracy of the pre-evaluation cannot be guaranteed.

[0007] The above method for pre-assessing wireless cell load based on the number of wireless cell users only considers the relationship between wireless cell load and the number of wireless cell users, and cannot guarantee the accuracy of wireless cell load pre-assessment.

[0008] There is currently no effective solution to the problem of low accuracy in load pre-assessment of the aforementioned wireless cells. Summary of the Invention

[0009] This invention provides a method, apparatus, and computer program product for pre-assessing the load of wireless cells, in order to at least solve the technical problem of low accuracy in pre-assessing the load of wireless cells.

[0010] According to one aspect of the present invention, a method for pre-assessing the load of a wireless cell is provided, comprising: acquiring a target measurement report and target operating parameter data of the wireless cell to be assessed; determining dynamic communication performance characteristics of the wireless cell to be assessed from the target measurement report; determining static communication performance characteristics of the wireless cell to be assessed from the target operating parameter data; and analyzing the dynamic and static communication performance characteristics using a cell load pre-assessment model to obtain a target load index of the wireless cell to be assessed, wherein the cell load pre-assessment model is pre-trained using multiple sets of training data through machine learning, and each set of training data includes: dynamic communication performance characteristics, static communication performance characteristics, and pre-calibrated sample load index of the same sample wireless cell.

[0011] Optionally, obtaining the target measurement report and target operating parameter data of the wireless cell to be evaluated includes: obtaining at least one historical measurement report of the wireless cell to be evaluated, wherein the historical measurement report is reported by a terminal accessing the wireless cell to be evaluated; analyzing at least one historical measurement report, and reallocating cell configuration information for each terminal according to a load balancing algorithm and a handover rule, wherein the load balancing algorithm is used to determine a first terminal that needs to be handed over from the wireless cell to be evaluated, and a second terminal that does not need to be handed over from the wireless cell to be evaluated, and the handover rule is used to allocate first cell configuration information to the first terminal and second cell configuration information to the second terminal, wherein the first cell configuration information is at least used to instruct the first terminal to hand over from the wireless cell to be evaluated to the target wireless cell, and the second cell configuration information is at least used to instruct the second terminal to re-access the wireless cell to be evaluated; receiving the target measurement report reported by the second terminal, and obtaining the target operating parameter data of the wireless cell to be evaluated.

[0012] Optionally, before analyzing the dynamic and static communication performance characteristics using a cell load pre-assessment model to obtain the target load index of the wireless cell to be evaluated, the method further includes: acquiring sample measurement reports, sample engineering parameter data, and sample performance index data of multiple sample wireless cells; determining the dynamic communication performance characteristics of the sample wireless cells from the sample measurement reports; determining the static communication performance characteristics of the sample wireless cells from the sample engineering parameter data; determining the sample load index of the sample wireless cells from the sample performance index data; and determining the dynamic communication performance characteristics, static communication performance characteristics, and sample load index of the same sample wireless cell as a set of training data.

[0013] Optionally, the method further includes: performing feature classification on the feature indicators in the dynamic communication performance characteristics and the static communication performance characteristics to obtain numerical features and classification features, wherein the numerical features include at least one of the feature indicators containing digital information, and the classification features include at least one of the feature indicators containing category information. The numerical features and the classification features are then combined using the cell load pre-assessment model to obtain load balancing features corresponding to the dynamic communication performance characteristics and the static communication performance characteristics.

[0014] Optionally, the load balancing features corresponding to the dynamic communication performance features and the static communication performance features are obtained by concatenating the numerical features and the classification features using the cell load pre-assessment model. This includes: processing the classification features using the embedded network layer in the cell load pre-assessment model to obtain a classification feature vector; processing the preset features using the normalized nonlinear feedforward neural network layer in the cell load pre-assessment model to obtain a preset feature vector, wherein the normalized nonlinear feedforward neural network layer is a first feedforward neural network layer, the preset features are the numerical features, and the preset feature vector is a numerical feature vector; and adding the classification feature vector and the numerical feature vector to obtain the load balancing features.

[0015] Optionally, after adding the classification feature vector and the numerical feature vector to obtain the load balancing feature, the method further includes: processing the preset feature using a normalized nonlinear feedforward neural network layer in the cell load pre-assessment model to obtain a preset feature vector, wherein the normalized nonlinear feedforward neural network layer is a second feedforward neural network layer, the preset feature is the load balancing feature, and the preset feature vector is a spatial feature vector; and analyzing the spatial feature vector using a linear adaptation network layer in the cell load pre-assessment model to obtain a preset load index, wherein the linear adaptation network layer is defined as follows: Linear_adaptation(*) is a linear adaptation function, * is the spatial feature vector of the preset radio cell, d is the feature dimension of the spatial feature vector, and w k Here, b is the weighting coefficient, b0 is the weighting bias, and the preset radio cell is either the radio cell to be evaluated or the sample radio cell. When the preset radio cell is the sample radio cell, the preset load index is the predicted load index obtained by pre-evaluating the sample radio cell. X i w is the spatial feature vector of the i-th sample wireless cell. k b0 and b0 are learnable parameters. When the preset radio cell is the radio cell to be evaluated, the preset load index is the target load index, w k b0 and b0 are trained in advance using the spatial feature vectors and sample load indices of the sample wireless cells.

[0016] Optionally, processing the preset features using the normalized nonlinear feedforward neural network layer in the cell load pre-assessment model to obtain the preset feature vector includes: processing the preset features using the normalization layer in the normalized nonlinear feedforward neural network layer to obtain a first feature vector, wherein the normalization layer is defined as follows: Normalization(*) is a normalization function, x is the preset feature of the preset wireless cell, E[x] is the expected mean of x, Var[x] is the expected variance of x, and ∈ is a pre-set hyperparameter. The preset wireless cell is the wireless cell to be evaluated or the sample wireless cell. When the preset wireless cell is the sample wireless cell, γ and β are learnable parameters, with γ initially set to 1 and β initially set to 0. When the preset wireless cell is the wireless cell to be evaluated, γ and β are pre-trained using the preset feature of the sample wireless cell and the sample load index. The first feature vector is processed by the linear network layer in the normalized nonlinear feedforward neural network layer to obtain the second feature vector, where the linear network layer is defined as follows: Linear(X) = X·A T+b, Linear(*) is a linear function, X is the first feature vector of the preset wireless cell, the preset wireless cell is the wireless cell to be evaluated or the sample wireless cell, when the preset wireless cell is the sample wireless cell, matrix A and vector b are learnable parameters, when the preset wireless cell is the wireless cell to be evaluated, matrix A and vector b are trained in advance using the first feature vector of the sample wireless cell and the sample load index, and the second feature vector is nonlinearly mapped using the activation function layer in the normalized nonlinear feedforward neural network layer to obtain the preset feature vector, wherein the activation function layer is defined as follows: Activation(x)=max(0,x)+a*min(0,x), Activation(*) is the activation function, x is the second feature vector of the preset wireless cell, the preset wireless cell is the wireless cell to be evaluated or the sample wireless cell, when the preset wireless cell is the sample wireless cell, a is a learnable parameter, when the preset wireless cell is the wireless cell to be evaluated, a is trained in advance using the first feature vector of the sample wireless cell and the sample load index.

[0017] Optionally, after classifying the feature indicators in the dynamic communication performance characteristics and the static communication performance characteristics to obtain numerical features and categorical features, the method further includes: preprocessing the numerical features using zero-mean unit variance normalization, wherein the definition of zero-mean unit variance normalization is as follows: x is the numerical feature, x mean x is the mean of the numerical features. std x is the standard deviation of the numerical characteristic. norm These are the preprocessed numerical features.

[0018] Optionally, the method further includes: training the cell load pre-assessment model using a mean squared error loss function, wherein the mean squared error loss function is defined as follows: MSE is the mean squared error, N is the number of sample wireless cells, and y i The sample load index is the pre-calibrated sample wireless cell for the i-th sample. The predicted load index obtained by using the cell load pre-assessment model to pre-assess the i-th sample wireless cell is determined by the smaller the mean square error calculated based on the mean square error loss function, indicating that the prediction accuracy of the cell load pre-assessment model is higher.

[0019] Optionally, the method further includes: iteratively training the cell load pre-assessment model using a momentum gradient descent algorithm, wherein the momentum gradient descent algorithm is defined as follows: θ t+1 =θ t -γg t θ t Let f(θ) be the parameter for the t-th iteration. t ) is the function in the t-th iteration. It is the differential function of the t-th iteration, g t This represents the gradient estimate for the t-th iteration. μ and γ are pre-set hyperparameters.

[0020] According to another aspect of the present invention, a wireless cell load pre-assessment apparatus is also provided, comprising: an acquisition module for acquiring a target measurement report of the wireless cell to be assessed and a target operating parameter data determination module for determining the dynamic communication performance characteristics of the wireless cell to be assessed from the target measurement report; a determination module for determining the static communication performance characteristics of the wireless cell to be assessed from the target operating parameter data; and an analysis module for analyzing the dynamic communication performance characteristics and static communication performance characteristics using a cell load pre-assessment model to obtain a target load index of the wireless cell to be assessed. The cell load pre-assessment model is pre-trained using multiple sets of training data through machine learning, and each set of training data includes: dynamic communication performance characteristics, static communication performance characteristics, and pre-calibrated sample load index of the same sample wireless cell.

[0021] According to another aspect of the present invention, a non-volatile storage medium is also provided, wherein a computer program is stored in the non-volatile storage medium, and the device where the non-volatile storage medium is located executes the above-mentioned load pre-assessment method for wireless cells by running the computer program.

[0022] According to another aspect of the present invention, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the above-described load pre-assessment method for wireless cells through the computer program.

[0023] According to another aspect of the present invention, a computer program product is also provided, including computer instructions that, when executed by a processor, implement the steps of the above-described load pre-assessment method for wireless cells.

[0024] In this embodiment of the invention, dynamic communication performance characteristics are determined based on the target measurement report of the wireless cell to be evaluated, and static communication performance characteristics are determined based on the target engineering parameter data of the wireless cell to be evaluated. Then, a cell load pre-evaluation model is used to pre-evaluate the dynamic and static communication performance characteristics of the wireless cell to be evaluated, thereby obtaining the target load index of the wireless cell to be evaluated. By utilizing the multi-dimensional characteristics affecting the load of the wireless cell for pre-evaluation, the load evaluation results of the wireless cell can be made more accurate, achieving the technical effect of improving the accuracy of the load pre-evaluation of the wireless cell, and thus solving the technical problem of low accuracy of the load pre-evaluation of the wireless cell. Attached Figure Description

[0025] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0026] Figure 1 This is a flowchart of a load pre-assessment method for a wireless cell according to an embodiment of the present invention;

[0027] Figure 2 This is a schematic diagram of a deep learning-based wireless cell load pre-assessment method according to an embodiment of the present invention;

[0028] Figure 3 This is a schematic diagram illustrating the reconstruction of historical measurement report data based on a load balancing algorithm according to an embodiment of the present invention;

[0029] Figure 4 This is a schematic diagram of a deep learning model for training wireless cell load pre-evaluation according to an embodiment of the present invention;

[0030] Figure 5 This is a schematic diagram of a deep learning model for wireless cell load pre-assessment according to an embodiment of the present invention;

[0031] Figure 6 This is a schematic diagram of a deep learning model for wireless cell load pre-assessment according to an embodiment of the present invention;

[0032] Figure 7 This is a schematic diagram of a deep learning-based wireless cell load pre-assessment device according to an embodiment of the present invention;

[0033] Figure 8 This is a schematic diagram comparing the pre-assessment error of a wireless cell load according to an embodiment of the present invention;

[0034] Figure 9 This is a schematic diagram comparing the proportion of high-load cells according to an embodiment of the present invention;

[0035] Figure 10 This is a schematic diagram comparing optimized cost inputs according to an embodiment of the present invention;

[0036] Figure 11 This is a schematic diagram of a load pre-assessment device for a wireless cell according to an embodiment of the present invention;

[0037] Figure 12 This is a structural block diagram of a computer terminal according to an embodiment of the present invention. Detailed Implementation

[0038] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0039] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0040] First, some nouns or terms that appear in the description of the embodiments of this application shall be interpreted as follows:

[0041] RSRP: Reference Signal Receiving Power, is a key parameter in mobile communication systems that represents the strength of wireless signals and is one of the physical layer measurement requirements. It is the average signal power received on all REs (resource particles) carrying the reference signal within a certain symbol.

[0042] According to an embodiment of the present invention, a method for pre-assessing the load of a wireless cell is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0043] Figure 1 This is a flowchart of a load pre-assessment method for a wireless cell according to an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes the following steps:

[0044] Step S102: Obtain the target measurement report and target engineering parameter data of the wireless cell to be evaluated;

[0045] Step S104: Determine the dynamic communication performance characteristics of the wireless cell to be evaluated from the target measurement report;

[0046] Step S106: Determine the static communication performance characteristics of the wireless cell to be evaluated from the target engineering parameter data;

[0047] Step S108: The cell load pre-assessment model is used to analyze the dynamic communication performance characteristics and static communication performance characteristics to obtain the target load index of the wireless cell to be assessed. The cell load pre-assessment model is trained in advance using multiple sets of training data through machine learning. Each set of training data includes: the dynamic communication performance characteristics, static communication performance characteristics and pre-calibrated sample load index of the same sample wireless cell.

[0048] In this embodiment of the invention, dynamic communication performance characteristics are determined based on the target measurement report of the wireless cell to be evaluated, and static communication performance characteristics are determined based on the target engineering parameter data of the wireless cell to be evaluated. Then, a cell load pre-evaluation model is used to pre-evaluate the dynamic and static communication performance characteristics of the wireless cell to be evaluated, thereby obtaining the target load index of the wireless cell to be evaluated. By utilizing the multi-dimensional characteristics affecting the load of the wireless cell for pre-evaluation, the load evaluation results of the wireless cell can be made more accurate, achieving the technical effect of improving the accuracy of the load pre-evaluation of the wireless cell, and thus solving the technical problem of low accuracy of the load pre-evaluation of the wireless cell.

[0049] In step S102 above, the wireless cell to be evaluated can be a serving cell that can provide communication services to the terminal.

[0050] In step S102 above, the target measurement report can be reported by the terminal, which is connected to the cell to be evaluated.

[0051] In step S102 above, the target technical parameters represent the relevant technical parameters of the wireless base station where the wireless cell to be evaluated is located, including the base station's location information, antenna azimuth angle, downtilt angle, transmit power, frequency band, bandwidth, coverage area, etc. This data is crucial for network planning, optimization, and maintenance, helping operators better manage and optimize wireless networks. Common technical parameters include the cell's base station ID, latitude and longitude, altitude, antenna azimuth angle, downtilt angle, coverage radius, coverage type, frequency band, power, etc.

[0052] In step S104 above, the dynamic communication performance characteristics include at least the number of users, traffic, rate, and coverage of the wireless cell to be evaluated per unit time.

[0053] Optionally, the unit of time includes, but is not limited to, hourly granularity and dayly granularity.

[0054] Optionally, the number of users includes, but is not limited to: the maximum number of users per unit time, the average number of users, and the minimum number of users.

[0055] Optionally, traffic includes, but is not limited to, the total traffic per unit time.

[0056] Optionally, the rate includes, but is not limited to: the maximum rate, the average rate, and the minimum rate per unit time.

[0057] Optionally, coverage includes, but is not limited to: the average RSRP value per unit time and weak coverage.

[0058] In step S106 above, the static communication performance characteristics include at least the following: the cell type, cell height, frequency, bandwidth and coverage scenario of the wireless cell to be evaluated.

[0059] In step S108 above, the target load indicators include, but are not limited to: the downlink PRB (Physical Resource Block) utilization rate and the uplink PRB utilization rate of the cell to be evaluated per unit time.

[0060] In step S108 above, the cell load pre-assessment model can be a neural network model, which can be loaded into memory.

[0061] For example, the raw data of a cell load pre-assessment model can be loaded from non-volatile memory into volatile memory so that the processor can run the cell load pre-assessment model. The raw data of the cell load pre-assessment model refers to unprocessed data, which typically includes the parameters and structural data of the cell load pre-assessment model. The structural data can be based on the computational relationships of the parameters, such as the forward propagation computational relationships between intermediate layers and between neurons. Specifically, the structural data can include the structure-related code of the cell load pre-assessment model, such as the code used to perform related calculations between intermediate layers and between neurons.

[0062] In one implementation, a region in memory can be partitioned for loading the cell load pre-assessment model, which may include a structure data storage region and a parameter storage region. The structure data storage region stores structure-related code, and the parameters referenced by it can be pointed to by pointers to the addresses of specific parameters in the parameter storage region. During the training process of the cell load pre-assessment model, it may be necessary to frequently update the parameters; in this case, updating the parameter values ​​in the parameter storage region is sufficient.

[0063] In the above embodiments of the present invention, the load of a wireless cell is not only related to the number of users in the wireless cell, but also to the traffic, rate, signal strength, etc. of the users in the wireless cell. The load pre-evaluation method of the wireless cell described in this application fully considers information such as traffic, rate, and signal strength, which can make the pre-evaluated target load indicators more accurate.

[0064] As an optional embodiment, obtaining the target measurement report and target operating parameter data of the wireless cell to be evaluated includes: obtaining at least one historical measurement report of the wireless cell to be evaluated, wherein the historical measurement report is reported by a terminal accessing the wireless cell to be evaluated; analyzing at least one historical measurement report and reallocating cell configuration information for each terminal according to a load balancing algorithm and a handover rule, wherein the load balancing algorithm is used to determine a first terminal that needs to be handed over from the wireless cell to be evaluated and a second terminal that does not need to be handed over from the wireless cell to be evaluated, and the handover rule is used to allocate first cell configuration information to the first terminal and second cell configuration information to the second terminal, wherein the first cell configuration information is used at least to instruct the first terminal to hand over from the wireless cell to be evaluated to the target wireless cell, and the second cell configuration information is used at least to instruct the second terminal to re-access the wireless cell to be evaluated; receiving the target measurement report reported by the second terminal and obtaining the target operating parameter data of the wireless cell to be evaluated.

[0065] In the embodiments described above, before pre-evaluating the load of the wireless cell to be evaluated, load balancing processing can be performed on the wireless cell to be evaluated. That is, the historical measurement reports of the wireless cell to be evaluated are analyzed to determine whether there is a load imbalance in the wireless cell to be evaluated, that is, to determine whether the wireless cell to be evaluated is overloaded. Then, if it is determined that the wireless cell to be evaluated is overloaded, the load balancing algorithm and handover rules are used to reallocate cell configuration information to each terminal accessing the wireless cell to be evaluated. When each terminal receives the cell configuration information, it can switch to the target wireless cell indicated by the cell configuration information, or it can re-access the wireless cell to be evaluated according to the reallocated cell configuration information, thereby completing the load balancing processing of the wireless cell to be evaluated. Then, when the wireless cell to be evaluated is under load balancing, the target measurement report and target operating parameter data of the wireless cell to be evaluated are obtained.

[0066] As an optional embodiment, before analyzing dynamic and static communication performance characteristics using a cell load pre-assessment model to obtain the target load index of the wireless cell to be assessed, the method further includes: acquiring sample measurement reports, sample engineering parameter data, and sample performance index data of multiple sample wireless cells; determining the dynamic communication performance characteristics of the sample wireless cells from the sample measurement reports; determining the static communication performance characteristics of the sample wireless cells from the sample engineering parameter data; determining the sample load index of the sample wireless cells from the sample performance index data; and determining the dynamic communication performance characteristics, static communication performance characteristics, and sample load index of the same sample wireless cell as a set of training data.

[0067] In the above embodiments of this application, the training data for the cell load pre-assessment model can be determined based on sample measurement reports, sample operating parameter data, and sample performance index data of multiple sample wireless cells. Based on the sample measurement report of each sample wireless cell, the dynamic communication performance characteristics of that sample wireless cell can be obtained. Based on the sample operating parameter data of each sample wireless cell, the static communication performance characteristics of that sample wireless cell can be obtained. Based on the sample performance index data of each sample wireless cell, the sample load index of that sample wireless cell can be obtained. Furthermore, by using the dynamic communication performance characteristics, static communication performance characteristics, and sample load index of the same sample wireless cell as a set of training data, multiple sets of training data determined based on multiple sample wireless cells can be obtained, thus realizing the determination of training data.

[0068] As an optional embodiment, the method further includes: performing feature classification on the feature indicators in the dynamic communication performance characteristics and static communication performance characteristics to obtain numerical features and classification features, wherein the numerical features include at least one feature indicator containing digital information, and the classification features include at least one feature indicator containing category information. The numerical features and classification features are then combined using a cell load pre-assessment model to obtain the load balancing features corresponding to the dynamic communication performance characteristics and static communication performance characteristics.

[0069] Optionally, dynamic communication performance characteristics may include at least one characteristic indicator, which may be the number of users, traffic, rate, or coverage of the wireless cell to be evaluated per unit time.

[0070] Optionally, static communication performance characteristics may include at least one characteristic indicator, which may be the site type, site height, frequency point, bandwidth, or coverage scenario of the wireless cell to be evaluated.

[0071] In the above embodiments of this application, the feature indicators in dynamic communication performance characteristics and static communication performance characteristics can be divided into numerical features and classification features. Different feature types require different processing and analysis methods. Therefore, after inputting dynamic communication performance characteristics and static communication performance characteristics into the cell load pre-assessment model, the cell load pre-assessment model can classify the feature indicators in dynamic communication performance characteristics and static communication performance characteristics, process the numerical features and classification features respectively, and then splice the processed numerical features and classification features to obtain load balancing features, thereby realizing the fusion of dynamic communication performance characteristics and static communication performance characteristics.

[0072] It should be noted that in both the training and usage phases of the cell load pre-assessment model, the model processes the input dynamic and static communication performance characteristics in the same way. The difference between the training and usage phases is as follows: In the usage phase, the cell load pre-assessment model outputs the target load index based on the input dynamic and static communication performance characteristics; in the training phase, the model outputs the predicted load index based on the input dynamic and static communication performance characteristics, and adjusts the parameters in the cell load pre-assessment model after outputting the predicted load index.

[0073] As an optional embodiment, the load balancing features corresponding to dynamic and static communication performance features are obtained by splicing numerical and classification features using a cell load pre-assessment model. This includes: processing the classification features using an embedded network layer in the cell load pre-assessment model to obtain a classification feature vector; processing the preset features using a normalized nonlinear feedforward neural network layer in the cell load pre-assessment model to obtain a preset feature vector, wherein the normalized nonlinear feedforward neural network layer is the first feedforward neural network layer, the preset features are numerical features, and the preset feature vector is a numerical feature vector; and adding the classification feature vector and the numerical feature vector to obtain the load balancing features.

[0074] In the above embodiments of this application, the cell load pre-assessment model can process numerical features and classification features in different ways. For example, the classification features can be processed using an embedded network layer, and the numerical features can be processed using a first feedforward neural network layer to obtain a classification feature vector determined based on the classification features and a numerical feature vector determined based on the numerical features. Then, the classification feature vector and the numerical feature vector are added together to obtain the load balancing features, thus realizing the fusion of numerical features and classification features.

[0075] As an optional embodiment, after adding the categorical feature vector and the numerical feature vector to obtain the load balancing feature, the method further includes: processing the preset features using a normalized nonlinear feedforward neural network layer in the cell load pre-assessment model to obtain a preset feature vector, wherein the normalized nonlinear feedforward neural network layer is a second feedforward neural network layer, the preset feature is the load balancing feature, and the preset feature vector is a spatial feature vector; and analyzing the spatial feature vector using a linear adaptation network layer in the cell load pre-assessment model to obtain a preset load index, wherein the linear adaptation network layer is defined as follows: Linear_adaptation(*) is a linear adaptation function, X is the spatial feature vector of the preset radio cell, d is the feature dimension of the spatial feature vector, and w k Here, b is the weighting coefficient, b0 is the weighting bias, and the preset radio cell is either the radio cell to be evaluated or the sample radio cell. If the preset radio cell is the sample radio cell, the preset load index is the predicted load index obtained from the pre-evaluation of the sample radio cell. X i w is the spatial feature vector of the i-th sample wireless cell. k b0 and b0 are learnable parameters. Given that the preset radio cell is the radio cell to be evaluated, the preset load metric is the target load metric, w k b0 and b0 are trained in advance using the spatial feature vectors and sample load indices of sample wireless cells.

[0076] In the above embodiments of this application, after fusing numerical features and classification features to obtain load balancing features, the first feedforward neural network layer can be used to process the load balancing features to obtain spatial feature vectors. Then, the linear adaptation network layer is used to analyze the spatial feature vectors to obtain preset load indicators, thereby realizing the pre-evaluation of preset load indicators using the cell load pre-evaluation model.

[0077] Optionally, if the dynamic and static communication performance characteristics of the input cell load pre-evaluation model are derived from the wireless cell to be evaluated, the preset load index output by the cell load pre-evaluation model is the target load index obtained by pre-evaluating the wireless cell to be evaluated.

[0078] Optionally, when the dynamic and static communication performance characteristics of the input cell load pre-evaluation model are derived from the sample wireless cell, the preset load index output by the cell load pre-evaluation model is the target load index obtained by pre-evaluating the wireless cell.

[0079] Optionally, both the first and second feedforward neural network layers can use the structure of a normalized nonlinear feedforward neural network layer.

[0080] As an optional embodiment, processing preset features using a normalized nonlinear feedforward neural network layer in the cell load pre-assessment model to obtain a preset feature vector includes: processing the preset features using a normalization layer in the normalized nonlinear feedforward neural network layer to obtain a first feature vector, wherein the normalization layer is defined as follows: Normalization(*) is the normalization function, x is the preset feature of the preset wireless cell, E[x] is the expected mean of x, Var[x] is the expected variance of x, ∈ is the preset hyperparameter, the preset wireless cell is the wireless cell to be evaluated or the sample wireless cell. When the preset wireless cell is the sample wireless cell, γ and β are learnable parameters, γ is initially 1, and β is initially 0. When the preset wireless cell is the wireless cell to be evaluated, γ and β are trained using the preset features and sample load index of the sample wireless cell. The first feature vector is processed by the linear network layer in the normalized nonlinear feedforward neural network layer to obtain the second feature vector, where the linear network layer is defined as follows: Linear(X) = X·A T+b, Linear(*) is a linear function, X is the first feature vector of the preset wireless cell, the preset wireless cell is the wireless cell to be evaluated or the sample wireless cell. When the preset wireless cell is the sample wireless cell, matrix A and vector b are learnable parameters. When the preset wireless cell is the wireless cell to be evaluated, matrix A and vector b are trained in advance using the first feature vector of the sample wireless cell and the sample load index. The second feature vector is nonlinearly mapped using the activation function layer in the normalized nonlinear feedforward neural network layer to obtain the preset feature vector. The activation function layer is defined as follows: Activation(x)=max(0,x)+a*min(0,x), Activation(*) is the activation function, x is the second feature vector of the preset wireless cell, the preset wireless cell is the wireless cell to be evaluated or the sample wireless cell. When the preset wireless cell is the sample wireless cell, a is a learnable parameter. When the preset wireless cell is the wireless cell to be evaluated, a is trained in advance using the first feature vector of the sample wireless cell and the sample load index.

[0081] In the above embodiments of this application, the normalized nonlinear feedforward neural network layer in the cell load pre-assessment model includes a normalization layer, a linear network layer, and an activation function layer. The normalization layer normalizes the preset features to obtain a first feature vector. Then, the linear network layer linearly processes the first feature vector to obtain a second feature vector. Finally, the activation function layer nonlinearly maps the second feature vector to obtain the preset feature vector. Thus, the normalized nonlinear feedforward neural network layer uses the normalization layer, the linear network layer, and the activation function layer to process the preset features.

[0082] As an optional embodiment, after classifying the feature indicators in the dynamic and static communication performance characteristics to obtain numerical features and categorical features, the method further includes: preprocessing the numerical features using zero-mean unit variance normalization, wherein zero-mean unit variance normalization is defined as follows: x is a numerical feature, x mean It is the mean of the numerical characteristics, x std It is the standard deviation of the numerical characteristic, x norm These are the preprocessed numerical features.

[0083] In the above embodiments of this application, zero-mean unit variance normalization is used for preprocessing of numerical features, thereby normalizing feature indicators from dynamic and static communication performance features.

[0084] As an optional embodiment, the method further includes: training a cell load pre-assessment model using a mean squared error loss function, wherein the mean squared error loss function is defined as follows: MSE is the mean squared error, N is the number of sample wireless cells, and y i For the i-th sample wireless cell, the pre-calibrated sample load index is... To obtain the predicted load index for the i-th sample wireless cell using the cell load pre-assessment model, the smaller the mean square error calculated based on the mean square error loss function, the higher the prediction accuracy of the cell load pre-assessment model.

[0085] In the above embodiments of this application, during the training phase of the cell load pre-assessment model, a predicted load index can be obtained by pre-assessing each sample wireless cell. The difference between the predicted load index of each sample wireless cell and the pre-calibrated sample load index of the sample wireless cell can be evaluated using the mean square error loss function. Then, the parameters in the cell load pre-assessment model can be adjusted according to the difference to achieve the training of the cell load pre-assessment model.

[0086] As an optional embodiment, the method further includes: iteratively training the cell load pre-assessment model using the momentum gradient descent algorithm, wherein the momentum gradient descent algorithm is defined as follows: θ t+1 =θ t -γg t θ t Let f(θ) be the parameter for the t-th iteration. t ) is the function in the t-th iteration. It is the differential function of the t-th iteration, g t This represents the gradient estimate for the t-th iteration. μ and γ are pre-set hyperparameters.

[0087] In the embodiments described above, the momentum gradient descent algorithm utilizes the concept of momentum to accelerate the gradient descent process, increasing the stability and convergence speed of the algorithm. This can improve the training speed of the cell load pre-assessment model, making the trained cell load pre-assessment model more accurate.

[0088] The present invention also provides an optional embodiment, which provides a deep learning-based method for pre-assessing wireless cell load, which can better support the effect of subsequent wireless cell load balancing and improve the efficiency of wireless cell load balancing.

[0089] Figure 2 This is a schematic diagram of a deep learning-based wireless cell load pre-assessment method according to an embodiment of the present invention, as shown below. Figure 2 As shown, the method includes the following steps:

[0090] Step S1: Reconstruct the historical measurement report based on the load balancing algorithm, and generate indicators such as the number of users, traffic, rate, and coverage after load balancing.

[0091] Step S2: Based on the target measurement report and target operating parameter data after load balancing, generate dynamic communication performance characteristics of the wireless cell, such as the number of users, traffic, rate, and coverage, and static communication performance characteristics, such as station type, station height, frequency, bandwidth, and coverage scenario, within a unit of time.

[0092] Step S3: Based on the trained deep learning model for wireless cell load pre-evaluation (i.e., cell load pre-evaluation model), input the dynamic communication performance characteristics and static communication performance characteristics after wireless cell load balancing, and output the load pre-evaluation results of the wireless cell (i.e., target load index).

[0093] The embodiments described above in this application, on the one hand, generate indicators including the number of users, traffic, rate, and coverage based on the load balancing algorithm, effectively increasing the feature dimension of load pre-assessment; on the other hand, they use deep learning algorithms to automatically model the high-order nonlinear relationship between wireless cell load and indicators such as the number of wireless cell users, traffic, rate, and coverage, improving the accuracy of wireless cell load pre-assessment. This allows for pre-assessment of the effectiveness of the load balancing method before its implementation, reducing the risk of network degradation that the load balancing method may cause, and improving the effect of load balancing optimization.

[0094] Figure 3 This is a schematic diagram illustrating the reconstruction of historical measurement report data based on a load balancing algorithm according to an embodiment of the present invention, such as... Figure 3 As shown, in step S1 above, reconstructing the historical measurement report data based on the load balancing algorithm includes the following steps:

[0095] Step S11: Based on the load balancing algorithm, reconstruct the balanced target measurement report.

[0096] Optionally, for each historical measurement report, the serving cell and neighboring cells, the frequency information of the serving cell and neighboring cells, the signal strength of the serving cell and neighboring cells, etc., are reassigned to the reporting terminal of each historical measurement report according to the load balancing algorithm and handover rules, and the frequency information, signal strength, etc. of the new serving cell and neighboring cells are adjusted at the same time.

[0097] Step S12: Based on the target measurement report, deduce and generate indicators such as the number of users, traffic, rate and coverage after load balancing (e.g., dynamic communication performance characteristics).

[0098] Optionally, based on the redistributed measurement report, the number of users, traffic, rate, and coverage (such as dynamic communication performance characteristics) after load balancing are recalculated according to the indicator definition.

[0099] In step S2 above, indicators such as the number of users, traffic, speed, and coverage of the wireless cell are generated based on the target measurement report within a unit of time.

[0100] Optionally, the unit of time includes, but is not limited to, hourly granularity and dayly granularity.

[0101] Optionally, the number of users includes, but is not limited to, the maximum number of users, the average number of users, and the minimum number of users.

[0102] Optionally, traffic includes, but is not limited to, total traffic.

[0103] Optionally, the rate of a wireless cell per unit time includes, but is not limited to, the maximum rate, the average rate, and the minimum rate.

[0104] Optionally, coverage includes, but is not limited to, average RSRP value, weak coverage, etc.

[0105] In step S2 above, the target engineering parameter data may optionally include, but is not limited to, station type, station height, frequency point, bandwidth, coverage scenario, etc.

[0106] Figure 4 This is a schematic diagram of a deep learning model for training wireless cell load pre-assessment according to an embodiment of the present invention, as shown below. Figure 4 As shown, before step S3 above, training samples (i.e., determining training data) can be constructed, and the training samples (i.e., determining training data) can be used to train the deep learning model for wireless cell load pre-evaluation (i.e., the cell load pre-evaluation model), including the following steps:

[0107] Step S311: Obtain the sample measurement report, sample engineering parameter data, and historical performance index data (i.e., sample performance index data) of the sample wireless cell to construct training samples.

[0108] Optionally, the sample operating parameters data include, but are not limited to: station type, station height, frequency, bandwidth, and coverage scenario.

[0109] Optionally, based on the sample measurement report, dynamic communication performance characteristics of the sample wireless cell, such as the number of users, traffic, rate, and coverage, can be calculated per unit time.

[0110] Optionally, the unit of time includes, but is not limited to, hourly granularity and dayly granularity.

[0111] Optionally, the number of users includes, but is not limited to: the maximum number of users, the average number of users, and the minimum number of users.

[0112] Optionally, traffic includes, but is not limited to, total traffic.

[0113] Optionally, the rate includes, but is not limited to: maximum rate, average rate, and minimum rate.

[0114] Optionally, coverage includes, but is not limited to, average RSRP value and weak coverage.

[0115] Optionally, based on historical performance data (i.e., sample performance data), the load index of the sample wireless cell per unit time (i.e., sample load index) can be calculated.

[0116] Optionally, the load indicators of the sample wireless cells (i.e., sample load indicators) include, but are not limited to: the downlink PRB (Physical Resource Block) utilization rate and the uplink PRB utilization rate of the wireless cells.

[0117] Optionally, based on the sample engineering parameter data of the sample wireless cell, static communication performance characteristics such as cell type, bandwidth, frequency, cell height, and coverage scenario of the sample wireless cell can be calculated.

[0118] As an optional implementation, the numerical features in the training data can be preprocessed using the zero-mean unit variance normalization method, which is defined as follows:

[0119]

[0120] x is a numerical feature, x mean It is the mean of the numerical characteristics, x std It is the standard deviation of the numerical characteristic.

[0121] In the above embodiment of the present invention, the preprocessed static communication performance characteristics, dynamic communication performance characteristics and tags (i.e. sample load indicators) in the training data are spliced ​​together to form a wireless cell load balancing training sample set.

[0122] Step S312: Construct and train a high-order nonlinear wireless cell load pre-evaluation deep learning model (i.e., cell load pre-evaluation model).

[0123] Optionally, after constructing the sample set, a deep learning algorithm can be used for training to generate a high-order nonlinear wireless cell load pre-evaluation deep learning model (i.e., a cell load pre-evaluation model).

[0124] Figure 5 This is a schematic diagram of a deep learning model for wireless cell load pre-assessment according to an embodiment of the present invention, as shown below. Figure 5As shown, the categorical features are processed by an embedding network layer to generate a d-dimensional categorical feature vector; the numerical features are processed by the first L-normalized nonlinear feedforward neural network layer (i.e., the first feedforward neural network layer) to generate a d-dimensional numerical feature vector; the sum of the d-dimensional categorical feature vector and the d-dimensional numerical feature vector is processed by the second L-normalized nonlinear feedforward neural network layer (i.e., the second feedforward neural network layer) to generate a d-dimensional spatial feature vector; the d-dimensional spatial feature vector is processed by a linear adaptation network layer to output the prediction result (i.e., the preset load index).

[0125] As an optional embodiment, the normalized nonlinear feedforward neural network includes: a normalization layer, a linear network layer, and an activation function layer.

[0126] As an optional example, the normalization layer is defined as follows: Where x is the preset feature of the input, E[x] is the expected mean of x, Var[x] is the expected variance of x, γ and β are learnable parameters, γ is initialized to 1, β is initialized to 0, and ∈ is a hyperparameter.

[0127] As an optional example, a linear network layer is defined as follows: Linear(X) = X·A T +b, where X is the input feature vector (such as the first feature vector), and matrix A and vector b are learnable parameters.

[0128] As an optional example, the activation function layer is defined as follows: Activation(x) = max(0,x) + a*min(0,x), where x is the input feature (such as the first feature vector) and a is a learnable parameter.

[0129] As an optional example, the Linear Adaptation network layer is defined as follows: Among them, X i X is the input feature vector (such as spatial feature vector) of the i-th sample wireless cell. i =[X i,1 ,X i,2 ,…,X i,d ], w k b0 and w are learnable parameters. k b is the weighting coefficient, and b0 is the weighting bias.

[0130] As an optional example, the high-order nonlinear wireless cell load pre-evaluation deep learning model (i.e., the cell load pre-evaluation model) uses mean-square error (MSE) as the loss function, which is defined as follows: Where N is the size of the sample set (i.e., the number of sample wireless cells), y i Let be the label of the i-th sample in the sample set (that is, the pre-defined sample load index of the i-th sample wireless cell). The mean square error (MSE) is the predicted value of the deep learning model for the i-th sample (i.e., the predicted load index obtained by pre-evaluating the i-th sample wireless cell). The smaller the MSE, the higher the prediction accuracy of the deep learning model.

[0131] As an optional example, the high-order nonlinear wireless cell load pre-assessment deep learning model (i.e., the cell load pre-assessment model) is trained using the momentum gradient descent algorithm, which is defined as follows: θ t+1 =θ t -γg t , where θ t Let f(θ) be the parameter for the t-th iteration. t ) is the function in the t-th iteration. It is the differential function of the t-th iteration. μ and γ are hyperparameters.

[0132] Figure 6 This is a schematic diagram of a deep learning model for wireless cell load pre-assessment according to an embodiment of the present invention, as shown below. Figure 6 As shown, the steps are as follows:

[0133] S321. Obtain the characteristic indicators of the wireless cell to be evaluated after load balancing (such as the dynamic communication performance characteristics and static communication performance characteristics of the wireless cell to be evaluated).

[0134] As an optional example, obtaining the static communication performance characteristics of the wireless cell to be evaluated after load balancing includes, but is not limited to: site type, site height, frequency point, bandwidth, and coverage scenario.

[0135] As an optional example, the dynamic communication performance characteristics of the wireless cell to be evaluated after load balancing include, but are not limited to: maximum number of users, average number of users, minimum number of users, total traffic, maximum rate, average rate, minimum rate, average RSRP value, and weak coverage.

[0136] As an optional embodiment, the dynamic and static communication performance characteristics of the wireless cell to be evaluated after load balancing are classified to obtain classification features and numerical features.

[0137] Optionally, classification features include, but are not limited to: site type, bandwidth, frequency point, and coverage scenario.

[0138] As an optional embodiment, the numerical characteristics of the wireless cell to be evaluated after load balancing are preprocessed using a zero-mean unit variance normalization method, wherein the zero-mean unit variance normalization method is defined as follows: x is a numerical feature, x mean It is the mean of the numerical characteristics, x std It is the standard deviation of the numerical characteristic.

[0139] Optionally, the preprocessed static communication performance characteristics and dynamic communication performance characteristics can be spliced ​​together to form the load balancing characteristics of the wireless cell to be evaluated.

[0140] S322. Based on the trained high-order nonlinear wireless cell load pre-evaluation deep learning model (i.e., cell load pre-evaluation model), output the wireless cell load pre-evaluation result (i.e., target load index).

[0141] As an optional example, the load balancing characteristics of the wireless cell to be evaluated are input into a trained high-order nonlinear wireless cell load pre-evaluation deep learning model (i.e., cell load pre-evaluation model), and the load prediction value of the wireless cell (i.e., target load index) is output.

[0142] Figure 7 This is a schematic diagram of a deep learning-based wireless cell load pre-assessment device according to an embodiment of the present invention, as shown below. Figure 7 As shown, it includes:

[0143] At least one transceiver is used to communicate between the wireless cell load pre-assessment server and the acquisition device and database, including receiving load pre-assessment instructions from the load pre-assessment server and selecting a wireless cell load pre-assessment model (i.e., cell load pre-assessment model) from the database and returning it to the load pre-assessment server; receiving the wireless cell's operating parameter data, measurement reports and historical performance index data from the acquisition device and transmitting them to the load pre-assessment server.

[0144] At least one processor and at least one memory communicatively connected to the processor, the memory storing program instructions executable by the processor, the program instructions invoked by the processor being able to execute the algorithms involved in the above steps, and the memory being used to store the corresponding results.

[0145] In terms of load pre-assessment features, compared with other load pre-assessment methods that only rely on the number of users to pre-assess wireless cell load indicators, the wireless cell load pre-assessment method of this application improves the accuracy of load indicator characterization by re-aggregating and calculating the measurement report data after load balancing, constructing multi-dimensional and multi-faceted features including the number of users, traffic, rate, and coverage.

[0146] In the above embodiments of this application, regarding the load pre-assessment method, compared to other load pre-assessment methods that assume a linear or third-order nonlinear relationship between the wireless cell load and the number of wireless cell users, the wireless cell load pre-assessment method of this invention does not impose linear or nonlinear constraints on the wireless cell load and characteristics such as the number of wireless cell users, traffic, rate, coverage, site type, bandwidth, frequency point, and site height. Instead, it uses a deep learning algorithm to model the nonlinear relationship between each characteristic and the wireless cell load, thereby improving the accuracy of load pre-assessment modeling.

[0147] Compared with two existing wireless cell load pre-assessment methods, this invention significantly reduces the error in wireless cell load pre-assessment, achieving more accurate wireless cell load pre-assessment.

[0148] Compared to technique 1, assuming a linear relationship between the wireless cell load and the number of wireless cell users, the number of wireless cell users after load balancing is first calculated, and then divided by the maximum user capacity of the wireless cell to pre-evaluate the wireless cell load.

[0149] Compared to technique 2, the number of wireless cell users after load balancing is first calculated, and then the wireless cell load is pre-evaluated based on the third-order polynomial fitting equation of wireless cell load and wireless cell user count.

[0150] Figure 8 This is a schematic diagram comparing the pre-assessment errors of a wireless cell load according to an embodiment of the present invention, as shown below. Figure 8 As shown, a wireless cell area was selected as a pilot project, and the mean absolute percentage error (MASE) was used to evaluate the pre-assessment accuracy of this application compared to comparative techniques 1 and 2. Theoretically, the smaller the MASE, the better the pre-assessment effect of the invention, and the higher its pre-assessment accuracy. This application significantly improves the accuracy of wireless cell load pre-assessment.

[0151] The load pre-assessment results of this application for wireless cells can be used for load balancing optimization of high-load cells in 4G / 5G networks. This can effectively save manpower and resource investment costs, improve network optimization efficiency, and can be promoted and applied throughout the province and even the country. It can not only reduce the proportion of high-load cells, but also reduce cost investment, resulting in considerable economic benefits.

[0152] Figure 9This is a schematic diagram comparing the proportion of high-load cells according to an embodiment of the present invention, such as... Figure 9 As shown, in this pilot area, the proportion of high-load cells has decreased significantly compared to the wireless cell load balancing optimization techniques of Comparative Technique 1 and Comparative Technique 2, based on the wireless cell load balancing optimization of this application.

[0153] Figure 10 This is a schematic diagram comparing optimized cost inputs according to an embodiment of the present invention, such as... Figure 10 As shown, the wireless cell load balancing optimization based on this application in the pilot area has significantly reduced costs compared to the wireless cell load balancing optimization technologies of Comparative Technology 1 and Comparative Technology 2.

[0154] The embodiments described above in this application, based on a load balancing algorithm, reconstruct historical measurement report data. This allows for the derivation of balanced wireless cell user numbers, traffic, speed, and coverage from historical measurement report sources, enabling more accurate estimation of these indicators and improving the accuracy of the load pre-assessment algorithm.

[0155] The embodiments described above in this application construct a sample set for wireless cell load pre-assessment based on sample engineering parameter data, sample measurement reports, and historical performance index data of wireless cells. This sample set is then trained using a deep learning algorithm to generate a high-order nonlinear deep learning model for wireless cell load pre-assessment (e.g., a cell load pre-assessment model). This method integrates measurement reports, engineering parameter data, and performance data to construct a high-order nonlinear wireless cell load pre-assessment model (e.g., a cell load pre-assessment model). Deep learning algorithms are used to model the nonlinear relationship between various features and wireless cell load, thereby improving the accuracy of load pre-assessment modeling.

[0156] In the above embodiments of this application, the pre-evaluation method comprehensively considers the dynamic communication performance characteristics of wireless cells, such as the number of users, traffic, rate, and coverage, while combining the static communication performance characteristics of wireless cell sites, such as site type, site height, frequency, bandwidth, and coverage scenario. It uses a wireless cell load pre-evaluation model (such as a cell load pre-evaluation model) to pre-evaluate the load indicators of the balanced wireless cell, which can accurately predict the load of the balanced wireless cell.

[0157] According to an embodiment of the present invention, a load pre-assessment device for a wireless cell is also provided. It should be noted that the load pre-assessment device for a wireless cell can be used to execute the load pre-assessment method for a wireless cell in the embodiments of the present invention, and the load pre-assessment method for a wireless cell in the embodiments of the present invention can be executed in the load pre-assessment device for a wireless cell.

[0158] Figure 11 This is a schematic diagram of a load pre-assessment device for a wireless cell according to an embodiment of the present invention, as shown below. Figure 11 As shown, the device may include: an acquisition module 1102, used to acquire a target measurement report and target operating parameter data of the wireless cell to be evaluated; a first determination module 1104, used to determine the dynamic communication performance characteristics of the wireless cell to be evaluated from the target measurement report; a second determination module 1106, used to determine the static communication performance characteristics of the wireless cell to be evaluated from the target operating parameter data; and an analysis module 1108, used to analyze the dynamic and static communication performance characteristics using a cell load pre-evaluation model to obtain the target load index of the wireless cell to be evaluated. The cell load pre-evaluation model is pre-trained using multiple sets of training data through machine learning. Each set of training data includes: the dynamic communication performance characteristics, static communication performance characteristics, and pre-calibrated sample load index of the same sample wireless cell.

[0159] It should be noted that the acquisition module 1102 in this embodiment can be used to execute step S102 in this application embodiment, the first determination module 1104 in this embodiment can be used to execute step S104 in this application embodiment, the second determination module 1106 in this embodiment can be used to execute step S106 in this application embodiment, and the analysis module 1108 in this embodiment can be used to execute step S108 in this application embodiment. The examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiments.

[0160] In this embodiment of the invention, dynamic communication performance characteristics are determined based on the target measurement report of the wireless cell to be evaluated, and static communication performance characteristics are determined based on the target engineering parameter data of the wireless cell to be evaluated. Then, a cell load pre-evaluation model is used to pre-evaluate the dynamic and static communication performance characteristics of the wireless cell to be evaluated, thereby obtaining the target load index of the wireless cell to be evaluated. By utilizing the multi-dimensional characteristics affecting the load of the wireless cell for pre-evaluation, the load evaluation results of the wireless cell can be made more accurate, achieving the technical effect of improving the accuracy of the load pre-evaluation of the wireless cell, and thus solving the technical problem of low accuracy of the load pre-evaluation of the wireless cell.

[0161] As an optional embodiment, the acquisition module includes: a first acquisition unit, configured to acquire at least one historical measurement report of the radio cell to be evaluated, wherein the historical measurement report is reported by a terminal accessing the radio cell to be evaluated; an analysis unit, configured to analyze at least one historical measurement report and reallocate cell configuration information for each terminal according to a load balancing algorithm and a handover rule, wherein the load balancing algorithm is used to determine a first terminal that needs to hand over from the radio cell to be evaluated and a second terminal that does not need to hand over from the radio cell to be evaluated, and the handover rule is used to allocate first cell configuration information to the first terminal and second cell configuration information to the second terminal, the first cell configuration information being used at least to instruct the first terminal to hand over from the radio cell to be evaluated to the target radio cell, and the second cell configuration information being used at least to instruct the second terminal to re-access the radio cell to be evaluated; and a second acquisition unit, configured to receive a target measurement report reported by the second terminal and acquire target operating parameter data of the radio cell to be evaluated.

[0162] As an optional embodiment, the apparatus further includes: an acquisition submodule, configured to acquire sample measurement reports, sample operating parameter data, and sample performance index data of multiple sample wireless cells before analyzing dynamic and static communication performance characteristics using a cell load pre-assessment model to obtain the target load index of the wireless cell to be evaluated; a first determination submodule, configured to determine the dynamic communication performance characteristics of the sample wireless cells from the sample measurement reports; a second determination submodule, configured to determine the static communication performance characteristics of the sample wireless cells from the sample operating parameter data; a third determination submodule, configured to determine the sample load index of the sample wireless cells from the sample performance index data; and a fourth determination submodule, configured to determine the dynamic communication performance characteristics, static communication performance characteristics, and sample load index of the same sample wireless cell as a set of training data.

[0163] As an optional embodiment, the apparatus further includes: a classification submodule, used to classify the feature indicators in the dynamic communication performance characteristics and static communication performance characteristics to obtain numerical features and classification features, wherein the numerical features include at least one feature indicator containing digital information, and the classification features include at least one feature indicator containing category information; and a splicing submodule, used to splice the numerical features and classification features using a cell load pre-assessment model to obtain load balancing features corresponding to the dynamic communication performance characteristics and static communication performance characteristics.

[0164] As an optional embodiment, the classification submodule includes: a first processing subunit, used to process classification features using an embedded network layer in the cell load pre-assessment model to obtain a classification feature vector; a second processing subunit, used to process preset features using a normalized nonlinear feedforward neural network layer in the cell load pre-assessment model to obtain a preset feature vector, wherein the normalized nonlinear feedforward neural network layer is the first feedforward neural network layer, the preset features are numerical features, and the preset feature vector is a numerical feature vector; and a third processing subunit, used to add the classification feature vector and the numerical feature vector to obtain load balancing features.

[0165] As an optional embodiment, the apparatus further includes: a fourth processing subunit, configured to, after adding the classification feature vector and the numerical feature vector to obtain the load balancing feature, process the preset feature using a normalized nonlinear feedforward neural network layer in the cell load pre-assessment model to obtain a preset feature vector, wherein the normalized nonlinear feedforward neural network layer is the second feedforward neural network layer, the preset feature is the load balancing feature, and the preset feature vector is a spatial feature vector; and an analysis subunit, configured to analyze the spatial feature vector using a linear adaptation network layer in the cell load pre-assessment model to obtain a preset load index, wherein the linear adaptation network layer is defined as follows: Linear_adaptation(*) is a linear adaptation function, X is the spatial feature vector of the preset radio cell, d is the feature dimension of the spatial feature vector, and w k Here, b is the weighting coefficient, b0 is the weighting bias, and the preset wireless cell is either the wireless cell to be evaluated or the sample wireless cell. When the preset wireless cell is the sample wireless cell, the preset load index is the predicted load index obtained by pre-evaluating the sample wireless cell using the cell load pre-evaluation model. X i w is the spatial feature vector of the i-th sample wireless cell. k b0 and b0 are learnable parameters. Given that the preset radio cell is the radio cell to be evaluated, the preset load metric is the target load metric, w k b0 and b0 are trained in advance using the spatial feature vectors and sample load indices of sample wireless cells.

[0166] As an optional embodiment, the third or fourth processing subunit includes: a fifth processing subunit, used to process preset features using a normalization layer in a normalized nonlinear feedforward neural network layer to obtain a first feature vector, wherein the normalization layer is defined as follows: Normalization(*) is the normalization function, x is the preset feature of the preset wireless cell, E[x] is the expected mean of x, Var[x] is the expected variance of x, ∈ is the preset hyperparameter, the preset wireless cell is the wireless cell to be evaluated or the sample wireless cell. When the preset wireless cell is the sample wireless cell, γ and β are learnable parameters, γ is initially 1 and β is initially 0. When the preset wireless cell is the wireless cell to be evaluated, γ and β are trained in advance using the preset features and sample load index of the sample wireless cell; The sixth processing subunit is used to process the first feature vector using the linear network layer in the normalized nonlinear feedforward neural network layer to obtain the second feature vector, where the linear network layer is defined as follows: Linear(X) = X·A T +b, Linear(*) is a linear function, X is the first feature vector of the preset wireless cell, the preset wireless cell is the wireless cell to be evaluated or the sample wireless cell. When the preset wireless cell is the sample wireless cell, matrix A and vector b are learnable parameters. When the preset wireless cell is the wireless cell to be evaluated, matrix A and vector b are trained in advance using the first feature vector of the sample wireless cell and the sample load index; The seventh processing subunit is used to perform nonlinear mapping on the second feature vector using the activation function layer in the normalized nonlinear feedforward neural network layer to obtain the preset feature vector. The activation function layer is defined as follows: Activation(x)=max(0,x)+a*min(0,x), Activation(*) is the activation function, x is the second feature vector of the preset wireless cell, the preset wireless cell is the wireless cell to be evaluated or the sample wireless cell. When the preset wireless cell is the sample wireless cell, a is a learnable parameter. When the preset wireless cell is the wireless cell to be evaluated, a is trained in advance using the first feature vector of the sample wireless cell and the sample load index.

[0167] As an optional embodiment, the apparatus further includes: a preprocessing submodule, used to preprocess the numerical features by classifying the feature indicators in the dynamic communication performance characteristics and static communication performance characteristics to obtain numerical features and classification features, and then using zero-mean unit variance normalization to preprocess the numerical features, wherein the definition of zero-mean unit variance normalization is as follows: x is a numerical feature, x mean It is the mean of the numerical characteristics, x std It is the standard deviation of the numerical characteristic, x norm These are the preprocessed numerical features.

[0168] As an optional embodiment, the apparatus further includes: a first training submodule, used to train a cell load pre-assessment model using a mean squared error loss function, wherein the mean squared error loss function is defined as follows: MSE is the mean squared error, N is the number of sample wireless cells, and y i For the i-th sample wireless cell, the pre-calibrated sample load index is... To obtain the predicted load index for the i-th sample wireless cell using the cell load pre-assessment model, the smaller the mean square error calculated based on the mean square error loss function, the higher the prediction accuracy of the cell load pre-assessment model.

[0169] As an optional embodiment, the apparatus further includes: a first training submodule, used to iteratively train the cell load pre-assessment model using a momentum gradient descent algorithm, wherein the momentum gradient descent algorithm is defined as follows: θ t+1 =θ t -γg t θ t Let f(θ) be the parameter for the t-th iteration. t ) is the function in the t-th iteration. It is the differential function of the t-th iteration, g t This represents the gradient estimate for the t-th iteration. μ and γ are pre-set hyperparameters.

[0170] Embodiments of the present invention may provide an electronic device that identifies a computer terminal, which may be any one of a group of computer terminals. Optionally, in this embodiment, the computer terminal may also be replaced by a mobile terminal or other terminal device.

[0171] Optionally, in this embodiment, the computer terminal may be located in at least one of a plurality of network devices in a computer network.

[0172] In this embodiment, the computer terminal described above can execute the program code for the following steps in the wireless cell load pre-assessment method: obtaining the target measurement report and target operating parameter data of the wireless cell to be assessed; determining the dynamic communication performance characteristics of the wireless cell to be assessed from the target measurement report; determining the static communication performance characteristics of the wireless cell to be assessed from the target operating parameter data; and analyzing the dynamic and static communication performance characteristics using a cell load pre-assessment model to obtain the target load index of the wireless cell to be assessed. The cell load pre-assessment model is pre-trained using multiple sets of training data through machine learning. Each set of training data includes: the dynamic communication performance characteristics, static communication performance characteristics, and pre-calibrated sample load index of the same sample wireless cell.

[0173] Figure 12 This is a structural block diagram of a computer terminal according to an embodiment of the present invention, such as... Figure 12As shown, the computer terminal 1200 may include one or more (only one is shown in the figure) processors 1202 and memory 1204.

[0174] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the wireless cell load pre-assessment method and apparatus in this embodiment of the invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the aforementioned wireless cell load pre-assessment method. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the terminal 1200 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0175] The processor can access the information and application programs stored in the memory via the transmission device to perform the following steps: acquiring the target measurement report and target operating parameter data of the wireless cell to be evaluated; determining the dynamic communication performance characteristics of the wireless cell to be evaluated from the target measurement report; determining the static communication performance characteristics of the wireless cell to be evaluated from the target operating parameter data; and analyzing the dynamic and static communication performance characteristics using a cell load pre-evaluation model to obtain the target load index of the wireless cell to be evaluated. The cell load pre-evaluation model is pre-trained using multiple sets of training data through machine learning. Each set of training data includes: the dynamic communication performance characteristics, static communication performance characteristics, and pre-calibrated sample load index of the same sample wireless cell.

[0176] Optionally, the processor may also execute program code for the following steps: acquiring at least one historical measurement report of the radio cell to be evaluated, wherein the historical measurement report is reported by a terminal accessing the radio cell to be evaluated; analyzing at least one historical measurement report and reallocating cell configuration information for each terminal according to a load balancing algorithm and a handover rule, wherein the load balancing algorithm is used to determine a first terminal that needs to be handed over from the radio cell to be evaluated and a second terminal that does not need to be handed over from the radio cell to be evaluated, and the handover rule is used to allocate first cell configuration information to the first terminal and second cell configuration information to the second terminal, wherein the first cell configuration information is used at least to instruct the first terminal to hand over from the radio cell to be evaluated to the target radio cell, and the second cell configuration information is used at least to instruct the second terminal to re-access the radio cell to be evaluated; receiving the target measurement report reported by the second terminal and acquiring the target operating parameter data of the radio cell to be evaluated.

[0177] Optionally, the processor may also execute program code that performs the following steps: acquiring sample measurement reports, sample operating parameter data, and sample performance index data of multiple sample wireless cells; determining the dynamic communication performance characteristics of the sample wireless cells from the sample measurement reports; determining the static communication performance characteristics of the sample wireless cells from the sample operating parameter data; determining the sample load index of the sample wireless cells from the sample performance index data; and determining the dynamic communication performance characteristics, static communication performance characteristics, and sample load index of the same sample wireless cell as a set of training data.

[0178] Optionally, the processor may also execute program code that performs the following steps: classifies the feature indicators in the dynamic communication performance characteristics and static communication performance characteristics to obtain numerical features and classification features, wherein the numerical features include at least one feature indicator containing digital information, and the classification features include at least one feature indicator containing category information. The numerical features and classification features are then combined using a cell load pre-assessment model to obtain the load balancing features corresponding to the dynamic communication performance characteristics and static communication performance characteristics.

[0179] Optionally, the processor may also execute program code for the following steps: processing the classification features using the embedded network layer in the cell load pre-assessment model to obtain a classification feature vector; processing the preset features using the normalized nonlinear feedforward neural network layer in the cell load pre-assessment model to obtain a preset feature vector, wherein the normalized nonlinear feedforward neural network layer is the first feedforward neural network layer, the preset features are numerical features, and the preset feature vector is a numerical feature vector; and adding the classification feature vector and the numerical feature vector to obtain the load balancing feature.

[0180] Optionally, the processor may also execute program code for the following steps: processing preset features using a normalized nonlinear feedforward neural network layer in the cell load pre-assessment model to obtain a preset feature vector, wherein the normalized nonlinear feedforward neural network layer is the second feedforward neural network layer, the preset features are load balancing features, and the preset feature vector is a spatial feature vector; analyzing the spatial feature vector using a linear adaptation network layer in the cell load pre-assessment model to obtain a preset load index, wherein the linear adaptation network layer is defined as follows: Linear_adaptation(*) is a linear adaptation function, X is the spatial feature vector of the preset radio cell, d is the feature dimension of the spatial feature vector, and w k Here, b is the weighting coefficient, b0 is the weighting bias, and the preset radio cell is either the radio cell to be evaluated or the sample radio cell. If the preset radio cell is the sample radio cell, the preset load index is the predicted load index obtained from the pre-evaluation of the sample radio cell. X i w is the spatial feature vector of the i-th sample wireless cell.k b0 and b0 are learnable parameters. Given that the preset radio cell is the radio cell to be evaluated, the preset load metric is the target load metric, w k b0 and b0 are trained in advance using the spatial feature vectors and sample load indices of sample wireless cells.

[0181] Optionally, the processor may also execute program code that performs the following steps: processing preset features using a normalization layer in a normalized nonlinear feedforward neural network to obtain a first feature vector, wherein the normalization layer is defined as follows: Normalization(*) is the normalization function, x is the preset feature of the preset wireless cell, E[x] is the expected mean of x, Var[x] is the expected variance of x, ∈ is the preset hyperparameter, the preset wireless cell is the wireless cell to be evaluated or the sample wireless cell. When the preset wireless cell is the sample wireless cell, γ and β are learnable parameters, γ is initially 1, and β is initially 0. When the preset wireless cell is the wireless cell to be evaluated, γ and β are trained using the preset features and sample load index of the sample wireless cell. The first feature vector is processed by the linear network layer in the normalized nonlinear feedforward neural network layer to obtain the second feature vector, where the linear network layer is defined as follows: Linear(X) = X·A T +b, Linear(*) is a linear function, X is the first feature vector of the preset wireless cell, the preset wireless cell is the wireless cell to be evaluated or the sample wireless cell. When the preset wireless cell is the sample wireless cell, matrix A and vector b are learnable parameters. When the preset wireless cell is the wireless cell to be evaluated, matrix A and vector b are trained in advance using the first feature vector of the sample wireless cell and the sample load index. The second feature vector is nonlinearly mapped using the activation function layer in the normalized nonlinear feedforward neural network layer to obtain the preset feature vector. The activation function layer is defined as follows: Activation(x)=max(0,x)+a*min(0,x), Activation(*) is the activation function, x is the second feature vector of the preset wireless cell, the preset wireless cell is the wireless cell to be evaluated or the sample wireless cell. When the preset wireless cell is the sample wireless cell, a is a learnable parameter. When the preset wireless cell is the wireless cell to be evaluated, a is trained in advance using the first feature vector of the sample wireless cell and the sample load index.

[0182] Optionally, the processor may also execute program code that performs the following steps: preprocessing the numerical features using zero-mean unit variance normalization, wherein zero-mean unit variance normalization is defined as follows: x is a numerical feature, x mean It is the mean of the numerical characteristics, xstd It is the standard deviation of the numerical characteristic, x norm These are the preprocessed numerical features.

[0183] Optionally, the processor may also execute program code that performs the following steps: training a cell load pre-assessment model using a mean squared error loss function, wherein the mean squared error loss function is defined as follows: MSE is the mean squared error, N is the number of sample wireless cells, and y i For the i-th sample wireless cell, the pre-calibrated sample load index is... To obtain the predicted load index for the i-th sample wireless cell using the cell load pre-assessment model, the smaller the mean square error calculated based on the mean square error loss function, the higher the prediction accuracy of the cell load pre-assessment model.

[0184] Optionally, the processor may also execute program code that performs the following steps: iteratively trains the cell load pre-assessment model using the momentum gradient descent algorithm, wherein the momentum gradient descent algorithm is defined as follows: θ t+1 =θ t -γg t θ t Let f(θ) be the parameter for the t-th iteration. t ) is the function in the t-th iteration. It is the differential function of the t-th iteration, g t This represents the gradient estimate for the t-th iteration. μ and γ are pre-set hyperparameters.

[0185] This invention provides a load pre-assessment scheme for wireless cells. Dynamic communication performance characteristics are determined based on the target measurement report of the wireless cell to be assessed, and static communication performance characteristics are determined based on the target engineering parameter data of the wireless cell to be assessed. Then, a cell load pre-assessment model is used to pre-assess the dynamic and static communication performance characteristics of the wireless cell to be assessed, obtaining the target load index of the wireless cell to be assessed. This utilizes multi-dimensional characteristics affecting the load of the wireless cell for pre-assessment, making the load assessment results more accurate and achieving the technical effect of improving the accuracy of wireless cell load pre-assessment. This solves the technical problem of low accuracy in wireless cell load pre-assessment.

[0186] Those skilled in the art will understand that Figure 12 The structure shown is for illustrative purposes only. The computer terminal can also be a smartphone (such as an Android phone, an iOS phone, etc.), a tablet computer, a mobile internet device (MID), a PAD, and other terminal devices. Figure 12 This does not limit the structure of the aforementioned electronic device. For example, the computer terminal 120 may also include components that are more advanced than those described above. Figure 12 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 12 The different configurations shown.

[0187] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a computer program instructing the hardware related to the terminal device. The computer program can be stored in a non-volatile medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc.

[0188] Embodiments of the present invention also provide a non-volatile storage medium. Optionally, in this embodiment, the aforementioned non-volatile storage medium can be used to store the program code executed by the wireless cell load pre-assessment method provided in the above embodiments.

[0189] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0190] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: obtaining the target measurement report and target operating parameter data of the wireless cell to be evaluated; determining the dynamic communication performance characteristics of the wireless cell to be evaluated from the target measurement report; determining the static communication performance characteristics of the wireless cell to be evaluated from the target operating parameter data; and analyzing the dynamic and static communication performance characteristics using a cell load pre-evaluation model to obtain the target load index of the wireless cell to be evaluated. The cell load pre-evaluation model is pre-trained using multiple sets of training data through machine learning. Each set of training data includes: the dynamic communication performance characteristics, static communication performance characteristics, and pre-calibrated sample load index of the same sample wireless cell.

[0191] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: obtaining at least one historical measurement report of the radio cell to be evaluated, wherein the historical measurement report is reported by a terminal accessing the radio cell to be evaluated; analyzing at least one historical measurement report and reallocating cell configuration information for each terminal according to a load balancing algorithm and a handover rule, wherein the load balancing algorithm is used to determine a first terminal that needs to be handed over from the radio cell to be evaluated and a second terminal that does not need to be handed over from the radio cell to be evaluated, and the handover rule is used to allocate first cell configuration information to the first terminal and second cell configuration information to the second terminal, wherein the first cell configuration information is at least used to instruct the first terminal to hand over from the radio cell to be evaluated to the target radio cell, and the second cell configuration information is at least used to instruct the second terminal to re-access the radio cell to be evaluated; receiving the target measurement report reported by the second terminal and obtaining the target operating parameter data of the radio cell to be evaluated.

[0192] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: acquiring sample measurement reports, sample operating parameter data, and sample performance index data of multiple sample wireless cells; determining the dynamic communication performance characteristics of the sample wireless cells from the sample measurement reports; determining the static communication performance characteristics of the sample wireless cells from the sample operating parameter data; determining the sample load index of the sample wireless cells from the sample performance index data; and determining the dynamic communication performance characteristics, static communication performance characteristics, and sample load index of the same sample wireless cell as a set of training data.

[0193] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: performing feature classification on the feature indicators in the dynamic communication performance characteristics and static communication performance characteristics to obtain numerical features and classification features, wherein the numerical features include at least one feature indicator containing digital information, and the classification features include at least one feature indicator containing category information. The numerical features and classification features are spliced ​​together using a cell load pre-assessment model to obtain the load balancing features corresponding to the dynamic communication performance characteristics and static communication performance characteristics.

[0194] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: processing classification features using an embedded network layer in the cell load pre-assessment model to obtain a classification feature vector; processing preset features using a normalized nonlinear feedforward neural network layer in the cell load pre-assessment model to obtain a preset feature vector, wherein the normalized nonlinear feedforward neural network layer is a first feedforward neural network layer, the preset features are numerical features, and the preset feature vector is a numerical feature vector; and adding the classification feature vector and the numerical feature vector to obtain the load balancing feature.

[0195] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: processing preset features using a normalized nonlinear feedforward neural network layer in the cell load pre-assessment model to obtain a preset feature vector, wherein the normalized nonlinear feedforward neural network layer is a second feedforward neural network layer, the preset features are load balancing features, and the preset feature vector is a spatial feature vector; analyzing the spatial feature vector using a linear adaptation network layer in the cell load pre-assessment model to obtain a preset load index, wherein the linear adaptation network layer is defined as follows: Linear_adaptation(*) is a linear adaptation function, X is the spatial feature vector of the preset radio cell, d is the feature dimension of the spatial feature vector, and w k Here, b is the weighting coefficient, b0 is the weighting bias, and the preset radio cell is either the radio cell to be evaluated or the sample radio cell. If the preset radio cell is the sample radio cell, the preset load index is the predicted load index obtained from the pre-evaluation of the sample radio cell. X i w is the spatial feature vector of the i-th sample wireless cell. k b0 and b0 are learnable parameters. Given that the preset radio cell is the radio cell to be evaluated, the preset load metric is the target load metric, w k b0 and b0 are trained in advance using the spatial feature vectors and sample load indices of sample wireless cells.

[0196] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: processing preset features using a normalization layer in a normalized nonlinear feedforward neural network layer to obtain a first feature vector, wherein the normalization layer is defined as follows: Normalization(*) is the normalization function, x is the preset feature of the preset wireless cell, E[x] is the expected mean of x, Var[x] is the expected variance of x, ∈ is the preset hyperparameter, the preset wireless cell is the wireless cell to be evaluated or the sample wireless cell. When the preset wireless cell is the sample wireless cell, γ and β are learnable parameters, γ is initially 1, and β is initially 0. When the preset wireless cell is the wireless cell to be evaluated, γ and β are trained using the preset features and sample load index of the sample wireless cell. The first feature vector is processed by the linear network layer in the normalized nonlinear feedforward neural network layer to obtain the second feature vector, where the linear network layer is defined as follows: Linear(X) = X·A T+b, Linear(*) is a linear function, X is the first feature vector of the preset wireless cell, the preset wireless cell is the wireless cell to be evaluated or the sample wireless cell. When the preset wireless cell is the sample wireless cell, matrix A and vector b are learnable parameters. When the preset wireless cell is the wireless cell to be evaluated, matrix A and vector b are trained in advance using the first feature vector of the sample wireless cell and the sample load index. The second feature vector is nonlinearly mapped using the activation function layer in the normalized nonlinear feedforward neural network layer to obtain the preset feature vector. The activation function layer is defined as follows: Activation(x)=max(0,x)+a*min(0,x), Activation(*) is the activation function, x is the second feature vector of the preset wireless cell, the preset wireless cell is the wireless cell to be evaluated or the sample wireless cell. When the preset wireless cell is the sample wireless cell, a is a learnable parameter. When the preset wireless cell is the wireless cell to be evaluated, a is trained in advance using the first feature vector of the sample wireless cell and the sample load index.

[0197] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: preprocessing the numerical features using zero-mean unit variance normalization, wherein zero-mean unit variance normalization is defined as follows: x is a numerical feature, x mean It is the mean of the numerical characteristics, x std It is the standard deviation of the numerical characteristic, x norm These are the preprocessed numerical features.

[0198] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: training a cell load pre-assessment model using a mean squared error loss function, wherein the mean squared error loss function is defined as follows: MSE is the mean squared error, N is the number of sample wireless cells, and y i For the i-th sample wireless cell, the pre-calibrated sample load index is... To obtain the predicted load index for the i-th sample wireless cell using the cell load pre-assessment model, the smaller the mean square error calculated based on the mean square error loss function, the higher the prediction accuracy of the cell load pre-assessment model.

[0199] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: iteratively training the cell load pre-assessment model using the momentum gradient descent algorithm, wherein the momentum gradient descent algorithm is defined as follows: θ t+1 =θ t -γgt θ t Let f(θ) be the parameter for the t-th iteration. t ) is the function in the t-th iteration. It is the differential function of the t-th iteration, g t This represents the gradient estimate for the t-th iteration. μ and γ are pre-set hyperparameters.

[0200] Embodiments of the present invention also provide a computer program product, including a computer program. Optionally, in this embodiment, when the computer program is executed by a processor, it implements the steps of the wireless cell load pre-assessment method provided in the above embodiments.

[0201] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0202] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0203] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0204] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0205] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0206] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a non-volatile storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a non-volatile storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned non-volatile storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0207] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for pre-assessing the load of a wireless cell, characterized in that, include: Obtain the target measurement report and target engineering parameter data of the wireless cell to be evaluated; From the target measurement report, determine the dynamic communication performance characteristics of the wireless cell to be evaluated; From the target engineering parameter data, determine the static communication performance characteristics of the wireless cell to be evaluated; The dynamic and static communication performance characteristics are analyzed using a cell load pre-assessment model to obtain the target load index of the wireless cell to be assessed. The cell load pre-assessment model is trained in advance using multiple sets of training data through machine learning. Each set of training data includes: dynamic communication performance characteristics, static communication performance characteristics, and pre-calibrated sample load index of the same sample wireless cell. The dynamic communication performance characteristics include at least the number of users, traffic, rate, and coverage indicators of the wireless cell to be evaluated per unit time, wherein the coverage indicators include the average reference signal received power (RSRP) and weak coverage rate per unit time; the static communication performance characteristics include at least the cell type, cell height, frequency, bandwidth, and coverage scenario of the wireless cell to be evaluated; the target load indicators include, but are not limited to, the downlink physical resource block utilization rate and the uplink physical resource block utilization rate of the wireless cell to be evaluated per unit time. The method further includes: The feature indicators in the dynamic communication performance characteristics and the static communication performance characteristics are classified using a cell load pre-assessment model to obtain numerical features and classification features. The numerical features include at least one feature indicator containing digital information, and the classification features include at least one feature indicator containing category information. By combining the numerical features and the classification features using the cell load pre-assessment model, load balancing features corresponding to the dynamic communication performance features and the static communication performance features are obtained. Based on the load balancing characteristics, a preset load index is obtained, wherein, when the dynamic communication performance characteristics and the static communication performance characteristics input to the cell load pre-evaluation model are derived from the wireless cell to be evaluated, the preset load index is the target load index.

2. The method according to claim 1, characterized in that, Obtaining the target measurement report and target engineering parameter data for the wireless cell to be evaluated includes: Obtain at least one historical measurement report of the wireless cell to be evaluated, wherein the historical measurement report is reported by a terminal accessing the wireless cell to be evaluated; Analyze at least one of the historical measurement reports and reallocate cell configuration information for each terminal according to a load balancing algorithm and a handover rule. The load balancing algorithm is used to determine a first terminal that needs to be handed over from the radio cell to be evaluated and a second terminal that does not need to be handed over from the radio cell to be evaluated. The handover rule is used to allocate first cell configuration information to the first terminal and second cell configuration information to the second terminal. The first cell configuration information is used at least to instruct the first terminal to hand over from the radio cell to be evaluated to the target radio cell, and the second cell configuration information is used at least to instruct the second terminal to re-access the radio cell to be evaluated. The system receives the target measurement report reported by the second terminal and obtains the target operating parameter data of the wireless cell to be evaluated.

3. The method according to claim 1, characterized in that, Before analyzing the dynamic and static communication performance characteristics using a cell load pre-assessment model to obtain the target load index of the wireless cell to be assessed, the method further includes: Obtain sample measurement reports, sample operating parameter data, and sample performance index data from multiple sample wireless cells; The dynamic communication performance characteristics of the sample wireless cell are determined from the sample measurement report; The static communication performance characteristics of the sample wireless cell are determined from the sample engineering parameter data; The sample load index of the sample wireless cell is determined from the sample performance index data. The dynamic communication performance characteristics, static communication performance characteristics, and sample load indicators of the same sample wireless cell are determined as a set of training data.

4. The method according to claim 1, characterized in that, By concatenating the numerical features and the classification features, the load balancing features corresponding to the dynamic communication performance features and the static communication performance features are obtained, including: The classification features are processed using the embedded network layer in the cell load pre-assessment model to obtain a classification feature vector; The preset features are processed by the normalized nonlinear feedforward neural network layer in the community load pre-assessment model to obtain the preset feature vector. The normalized nonlinear feedforward neural network layer is the first feedforward neural network layer, the preset features are the numerical features, and the preset feature vector is the numerical feature vector. The load balancing feature is obtained by adding the classification feature vector and the numerical feature vector.

5. The method according to claim 4, characterized in that, After adding the classification feature vector and the numerical feature vector to obtain the load balancing feature, the method further includes: The preset features are processed by the normalized nonlinear feedforward neural network layer in the community load pre-assessment model to obtain a preset feature vector. The normalized nonlinear feedforward neural network layer is the second feedforward neural network layer, the preset features are the load balancing features, and the preset feature vector is a spatial feature vector. The spatial feature vector is analyzed using the linear adaptation network layer in the aforementioned cell load pre-assessment model to obtain a preset load index. The linear adaptation network layer is defined as follows: , It is a linear fit function. It is the spatial feature vector of the preset wireless cell, where d is the feature dimension of the spatial feature vector. These are the weighting coefficients. As a weighting bias, the preset radio cell is either the radio cell to be evaluated or the sample radio cell. If the preset radio cell is the sample radio cell, the preset load index is a predicted load index obtained by pre-evaluating the sample radio cell. It is the first The spatial feature vectors of the sample wireless cells , and These are learnable parameters. When the preset wireless cell is the wireless cell to be evaluated, the preset load index is the target load index. and It is trained in advance using the spatial feature vectors and sample load indices of the sample wireless cells.

6. The method according to claim 4 or 5, characterized in that, The preset features are processed using the normalized nonlinear feedforward neural network layer in the aforementioned community load pre-assessment model to obtain a preset feature vector, which includes: The preset features are processed using the normalization layer in the normalized nonlinear feedforward neural network to obtain a first feature vector, wherein the normalization layer is defined as follows: , It is a normalization function. The preset features of the preset wireless cell are as described above. yes The expected mean, yes Expected variance These are pre-set hyperparameters. The preset radio cell is either the radio cell to be evaluated or the sample radio cell. In the case where the preset radio cell is the sample radio cell... and These are learnable parameters. The initial value is 1. The initial value is 0, assuming the preset radio cell is the radio cell to be evaluated. and It is trained in advance using the preset features and sample load indicators of the sample wireless cells; The first feature vector is processed using a linear network layer in the normalized nonlinear feedforward neural network layer to obtain a second feature vector, wherein the linear network layer is defined as follows: , It is a linear function. It is the first feature vector of the preset wireless cell, where the preset wireless cell is the wireless cell to be evaluated or the sample wireless cell. When the preset wireless cell is the sample wireless cell, the matrix... sum vector These are learnable parameters; given the preset radio cell as the radio cell to be evaluated, the matrix... sum vector It is trained in advance using the first feature vector of the sample wireless cell and the sample load index; The second feature vector is nonlinearly mapped using the activation function layer in the normalized nonlinear feedforward neural network layer to obtain the preset feature vector, wherein the activation function layer is defined as follows: , It is an activation function. It is the second feature vector of the preset wireless cell, where the preset wireless cell is either the wireless cell to be evaluated or the sample wireless cell. In the case where the preset wireless cell is the sample wireless cell... These are learnable parameters, assuming the preset radio cell is the radio cell to be evaluated. It is trained in advance using the first feature vector of the sample wireless cell and the sample load index.

7. The method according to claim 1, characterized in that, After classifying the feature indicators in the dynamic communication performance characteristics and the static communication performance characteristics to obtain numerical features and categorical features, the method further includes: The numerical features are preprocessed using zero-mean unit variance normalization, where zero-mean unit variance normalization is defined as follows: , It is the numerical feature, It is the mean of the numerical features. It is the standard deviation of the numerical feature. These are the preprocessed numerical features.

8. The method according to claim 1, characterized in that, The method further includes: The cell load pre-assessment model is trained using a mean squared error loss function, wherein the mean squared error loss function is defined as follows: , It is the mean square error. The number of sample wireless cells. For the first The sample load index of each of the sample wireless cells is pre-calibrated. To utilize the aforementioned community load pre-assessment model for the first The predicted load index obtained from the pre-evaluation of the sample wireless cells is such that the smaller the mean square error calculated based on the mean square error loss function, the higher the prediction accuracy of the cell load pre-evaluation model.

9. The method according to claim 1, characterized in that, The method further includes: The momentum gradient descent algorithm is used to iteratively train the cell load pre-assessment model, wherein the momentum gradient descent algorithm is defined as follows: , , For the first The parameters for the next iteration It is the first The function of the nth iteration It is the first The differential function of the next iteration Indicates the first The gradient estimate of the next iteration. , and These are pre-set hyperparameters.

10. A load pre-assessment device for a wireless cell, characterized in that, include: The acquisition module is used to acquire the target measurement report and target engineering parameter data of the wireless cell to be evaluated; The first determining module is used to determine the dynamic communication performance characteristics of the wireless cell to be evaluated from the target measurement report; The second determining module is used to determine the static communication performance characteristics of the wireless cell to be evaluated from the target engineering parameter data; The analysis module is used to analyze the dynamic communication performance characteristics and static communication performance characteristics using a cell load pre-assessment model to obtain the target load index of the wireless cell to be evaluated. The cell load pre-assessment model is trained in advance using multiple sets of training data through machine learning. Each set of training data includes: dynamic communication performance characteristics, static communication performance characteristics, and pre-calibrated sample load index of the same sample wireless cell. The dynamic communication performance characteristics include at least the number of users, traffic, rate, and coverage indicators of the wireless cell to be evaluated per unit time, wherein the coverage indicators include the average reference signal received power (RSRP) and weak coverage rate per unit time; the static communication performance characteristics include at least the cell type, cell height, frequency, bandwidth, and coverage scenario of the wireless cell to be evaluated; the target load indicators include, but are not limited to, the downlink physical resource block utilization rate and the uplink physical resource block utilization rate of the wireless cell to be evaluated per unit time. The device is also used for The feature indicators in the dynamic communication performance characteristics and the static communication performance characteristics are classified using a cell load pre-assessment model to obtain numerical features and classification features. The numerical features include at least one feature indicator containing digital information, and the classification features include at least one feature indicator containing category information. By combining the numerical features and the classification features using the cell load pre-assessment model, load balancing features corresponding to the dynamic communication performance features and the static communication performance features are obtained. Based on the load balancing characteristics, a preset load index is obtained, wherein, when the dynamic communication performance characteristics and the static communication performance characteristics input to the cell load pre-evaluation model are derived from the wireless cell to be evaluated, the preset load index is the target load index.

11. A non-volatile storage medium, characterized in that, The non-volatile storage medium stores a computer program, wherein the device containing the non-volatile storage medium executes the load pre-assessment method for the wireless cell according to any one of claims 1 to 9 by running the computer program.

12. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to execute the load pre-assessment method for the wireless cell according to any one of claims 1 to 9 through the computer program.

13. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the load pre-assessment method for the wireless cell according to any one of claims 1 to 9.

Images