Cell parameter configuration method and apparatus, and computing device cluster

Abstract

The cell parameter configuration method comprises the steps that first traffic statistic data of a first cell in a target office point is acquired, and the target office point is a wireless cellular network belonging to the same operator in a region; based on the first traffic statistic data, performing traffic scene classification on the first cell to obtain scene classification information of the first cell, the scene classification information being used for indicating the probability that the first cell belongs to each traffic scene in the at least one traffic scene; and determining an optimal configuration value of the to-be-optimized parameter in the first cell based on the first traffic statistic data and the scene classification information. Therefore, when parameter configuration is carried out on the cells in the office point, telephone traffic scene division is carried out on the cells through the real telephone traffic data in the cells, and configuration of the parameters to be optimized in the cells is determined through the telephone traffic scenes obtained through division, so that differentiated parameter recommendation can be realized for the cells of different telephone traffic scenes, and the parameter optimization effect is improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence (AI) technology, and in particular to a method, apparatus and computing device cluster for configuring cell parameters. Background Technology

[0002] With the continuous development of wireless communication technology, mobile communication has become an indispensable part of people's daily lives. In mobile communication infrastructure, the wireless cellular network plays a crucial role. It is built by mobile operators and provides users with mobile data and voice services. A wireless cellular network consists of a large number of wireless base stations, each emitting a signal that covers a specific surrounding area. Users can communicate with base stations through their mobile terminals to obtain data (such as accessing the internet) or voice services. When users move between different areas, their mobile terminals can switch to different base stations to obtain uninterrupted mobile communication services. The area covered by each wireless base station can be divided into multiple wireless cells based on geographical location or signal frequency band. Each wireless cell can be considered an independent network control unit. Each wireless cell has a large number of parameters that need to be configured. These parameter configurations are crucial to the quality of service of the wireless cell, determining many characteristics of the cell, such as coverage area, communication method with user terminals, and handover control between cells. Therefore, mobile operators need to continuously optimize the parameter configuration of wireless cells to ensure a good user experience and meet the needs of a wide range of users. Currently, the parameter optimization process for wireless cells mainly relies on the manual experience of experts, resulting in poor optimization effects. Summary of the Invention

[0003] This application provides a method, apparatus, computing device cluster, computer storage medium, and computer product for configuring cell parameters, which can improve the effect of cell parameter optimization.

[0004] Firstly, this application provides a cell parameter configuration method, comprising: acquiring first call statistics data of a first cell in a target location, wherein the target location is a wireless cellular network belonging to the same operator in a geographical area; classifying the first cell into call scenarios based on the first call statistics data to obtain scenario classification information of the first cell, wherein the scenario classification information is used to indicate the probability that the first cell belongs to each of at least one call scenario; and determining the optimal configuration value of the parameters to be optimized in the first cell based on the first call statistics data and the scenario classification information. For example, the first call statistics data may include: state data of the first cell, such as: the average number of users per unit time period, the average number of active users per unit time period, the average length of data packets per unit time period, etc. Furthermore, the scenario classification information may, but is not limited to, be obtained by classifying the first call statistics data through a neural network, i.e., the call scenarios are learned by a neural network model.

[0005] In this way, when configuring parameters in a cell at a central office location, the cell is divided into traffic scenarios based on real call statistics data within the cell. The configuration of parameters to be optimized in the cell is determined by the traffic scenarios obtained from the division. This allows for differentiated parameter recommendations for cells with different traffic scenarios, thereby improving the parameter optimization effect.

[0006] In one possible implementation, there are multiple parameters to be optimized, which are divided into multiple parameter groups. The traffic scenario is obtained by treating these multiple parameter groups as a whole, and the number of scenario classification information is 1. In this way, the traffic scenario is divided for the first cell under all parameter groups, which improves the efficiency of scenario division.

[0007] In one possible implementation, the scene classification information is an M-dimensional vector, where M is the number of call scenarios, each element in the M-dimensional vector is a value between [0,1], and the sum of all elements is 1. The value of an element in the M-dimensional vector represents the probability that the first cell belongs to a call scenario.

[0008] In one possible implementation, there are multiple parameters to be optimized, which are divided into multiple parameter groups. Each parameter group is further divided into at least one traffic scenario. The scenario classification information includes multiple sub-scenario classification information, the number of which is the same as the number of parameter groups. Each sub-scenario classification information indicates the probability that the first cell belongs to any of the traffic scenarios defined within a parameter group. This allows for traffic scenario classification of the first cell within each parameter group, resulting in finer-grained scenario classification and improved accuracy of parameter configuration.

[0009] In one possible implementation, multiple parameter sets include: a first parameter set and a second parameter set. Scene classification information includes: a first sub-scene classification information and a second sub-scene classification information. The first sub-scene classification information is an S-dimensional vector, where S represents the number of traffic scenes divided under the first parameter set. The second sub-scene classification information is a Q-dimensional vector, where Q represents the number of traffic scenes divided under the second parameter set. S and Q may be equal or unequal. Each element in the S-dimensional vector is a value between [0,1], and the sum of all elements is 1. The value of an element in the S-dimensional vector represents the probability that the first cell belongs to a traffic scene. Similarly, each element in the Q-dimensional vector is a value between [0,1], and the sum of all elements is 1. The value of an element in the Q-dimensional vector represents the probability that the first cell belongs to a traffic scene.

[0010] In one possible implementation, based on the first call statistics data, the first cell is classified into call scenarios to obtain scenario classification information for the first cell. This includes: extracting features from the first call statistics data to obtain a cell representation of the first cell; and calculating the scenario classification information of the first cell based on the cell representation. In this way, the call scenarios of the cell can be divided using the cell representation. For example, before extracting features from the first call statistics data, redundant data can be removed to improve the accuracy of subsequent call scenario classification.

[0011] In one possible implementation, there are multiple parameters to be optimized, which are divided into multiple parameter groups. Then, based on the first call statistics data and scene classification information, the optimal configuration values ​​for the parameters to be optimized in the first cell are determined. This includes: iteratively searching for configuration values ​​for the parameters to be optimized; and, in each iteration, calculating the global performance contributed by each parameter group in that iteration based on the first call statistics data, scene classification information, and the configuration values ​​of the parameters to be optimized found in that iteration; and, based on the global performance of each parameter group calculated in each iteration, selecting the optimal configuration value from the configuration values ​​of the parameters to be optimized found in the iterative search, where the global performance contributed by each parameter group is optimal under the optimal configuration value. In this way, by conducting multiple parameter searches and calculating the global performance contributed by each parameter group under each searched parameter value, the optimal parameter configuration value can be selected, improving the accuracy of parameter configuration.

[0012] In one possible implementation, based on the first set of statistical data, scene classification information, and the configuration values ​​of the parameters to be optimized found in the current iteration, the global performance contributed by each parameter group in the current iteration is calculated. This includes: calculating the performance contribution ratio of each parameter group based on the configuration values ​​of each parameter to be optimized in each parameter group found in the current iteration and the scene classification information; calculating the weight of the performance contribution ratio of each parameter group based on the first set of statistical data; and calculating the global performance contributed by each parameter group in the current iteration based on the performance contribution ratio of each parameter group and its corresponding weight. Since the performance contribution ratio of each parameter group may be different under different parameter configurations, weighting the performance contribution ratio of each parameter group can better reflect the global performance contributed by each parameter group, thereby improving the accuracy of parameter configuration.

[0013] In one possible implementation, before classifying the traffic scenarios of the first cell based on the first call statistics data, the method further includes: selecting neural network modules related to the parameter configuration of the first cell from a model library containing multiple neural network modules, based on the configuration requirements of the target site. The neural network modules in the model library are pre-trained or some are initialized. At least the selected neural network modules related to the parameter configuration of the first cell are assembled to obtain a cell parameter configuration model. This model is used to process at least the first call statistics data to obtain the optimal configuration values ​​for the parameters to be optimized in the first cell. Since the neural network modules in the model library are pre-trained or some are initialized, the cell parameter configuration model can be customized based on the site's configuration requirements, making model deployment more flexible and eliminating the need for re-pre-training for each site, thus reducing deployment costs. Furthermore, since the modules in the assembled cell parameter configuration model are pre-trained or some are initialized, they can be directly recommended in new sites, or only require a small amount of data for model fine-tuning, significantly shortening the optimization cycle for new sites.

[0014] One possible implementation also includes fine-tuning the cell parameter configuration model. The data required for fine-tuning is collected from the target site after assigning differentiated parameter configuration values ​​to cells with similar call scenarios. Since cells with similar call scenarios use different parameter configuration values, collecting differentiated data from these cells increases the diversity of the fine-tuning samples, thereby improving the model's generalization ability, reducing overfitting, and enhancing the fine-tuning effect.

[0015] Secondly, this application provides a cell parameter configuration device, including: an acquisition module and a processing module. The acquisition module is used to acquire first call statistics data of a first cell in a target location, where the target location is a wireless cellular network belonging to the same operator within a geographical area. The processing module is used to classify the first cell into call scenarios based on the first call statistics data to obtain scenario classification information for the first cell, the scenario classification information indicating the probability that the first cell belongs to each of at least one call scenario; and to determine the optimal configuration value of the parameters to be optimized in the first cell based on the first call statistics data and the scenario classification information.

[0016] In one possible implementation, there are multiple parameters to be optimized, which are divided into multiple parameter groups. The call scenario is divided by treating the multiple parameter groups as a whole, and the number of scenario classification information is 1.

[0017] In one possible implementation, the scene classification information is an M-dimensional vector, where M is the number of call scenarios, each element in the M-dimensional vector is a value between [0,1], and the sum of all elements is 1. The value of an element in the M-dimensional vector represents the probability that the first cell belongs to a call scenario.

[0018] In one possible implementation, there are multiple parameters to be optimized, which are divided into multiple parameter groups. Each parameter group is further divided into at least one traffic scenario. The scenario classification information includes multiple sub-scenario classification information, the number of which is the same as the number of parameter groups. A sub-scenario classification information indicates the probability that the first cell belongs to any of the traffic scenarios defined within a parameter group.

[0019] In one possible implementation, multiple parameter sets include: a first parameter set and a second parameter set. Scene classification information includes: a first sub-scene classification information and a second sub-scene classification information. The first sub-scene classification information is an S-dimensional vector, where S represents the number of traffic scenes divided under the first parameter set. The second sub-scene classification information is a Q-dimensional vector, where Q represents the number of traffic scenes divided under the second parameter set. S and Q may be equal or unequal. Each element in the S-dimensional vector is a value between [0,1], and the sum of all elements is 1. The value of an element in the S-dimensional vector represents the probability that the first cell belongs to a traffic scene. Similarly, each element in the Q-dimensional vector is a value between [0,1], and the sum of all elements is 1. The value of an element in the Q-dimensional vector represents the probability that the first cell belongs to a traffic scene.

[0020] In one possible implementation, when the processing module performs call scenario classification on the first cell based on the first call statistics data to obtain the scenario classification information of the first cell, it specifically performs the following: extracts features from the first call statistics data to obtain the cell representation of the first cell; and calculates the scenario classification information based on the cell representation of the first cell.

[0021] In one possible implementation, there are multiple parameters to be optimized, which are divided into multiple parameter groups. In this case, when the processing module determines the optimal configuration values ​​of the parameters to be optimized in the first cell based on the first call statistics data and scene classification information, it specifically performs the following: iteratively searching for configuration values ​​of the parameters to be optimized; and in each iteration, calculating the global performance contributed by each parameter group in the current iteration based on the first call statistics data, scene classification information, and the configuration values ​​of the parameters to be optimized found in the current iteration; and based on the global performance contributed by each parameter group calculated in each iteration, selecting the optimal configuration value from the configuration values ​​of the parameters to be optimized found in the iterative search, where the global performance contributed by each parameter group is optimal under the optimal configuration value.

[0022] In one possible implementation, when the processing module calculates the global performance contributed by each parameter group in the current iteration based on the first call statistics data, scene classification information, and the configuration values ​​of the parameters to be optimized found in the current iteration, it specifically performs the following: calculating the performance contribution ratio of each parameter group based on the configuration values ​​of each parameter to be optimized in each parameter group found in the current iteration and the scene classification information; calculating the weight of the performance contribution ratio of each parameter group based on the first call statistics data; and calculating the global performance contributed by each parameter group in the current iteration based on the performance contribution ratio of each parameter group and the corresponding weight.

[0023] In one possible implementation, before classifying the traffic scenarios of the first cell based on the first call statistics data, the processing module is further configured to: based on the configuration requirements of the target site, select neural network modules related to the parameter configuration of the first cell from a model library containing multiple neural network modules, wherein the neural network modules in the model library are pre-trained or initialized; and assemble at least the selected neural network modules related to the parameter configuration of the first cell to obtain a cell parameter configuration model, wherein the cell parameter configuration model is used to process at least the first call statistics data to obtain the optimal configuration values ​​of the parameters to be optimized in the first cell.

[0024] In one possible implementation, the processing module is further configured to: fine-tune the cell parameter configuration model, wherein the data required for fine-tuning is: collected from the target site after assigning differentiated parameter configuration values ​​to cells with similar call scenarios in the target site.

[0025] Thirdly, this application provides a cell parameter configuration apparatus, including at least one processor and an interface; the at least one processor acquires program instructions through the interface; the at least one processor is used to execute program line instructions to implement the method described in the first aspect or any possible implementation of the first aspect. Exemplarily, the cell parameter configuration apparatus may be, but is not limited to, a chip.

[0026] Fourthly, this application provides a computing device cluster, including at least one computing device, each computing device including a processor and a memory; the processor of the at least one computing device is used to execute instructions stored in the memory of the at least one computing device, so that the computing device cluster performs the method described in the first aspect or any possible implementation of the first aspect.

[0027] Fifthly, this application provides a computer-readable storage medium including computer program instructions that, when executed by a cluster of computing devices, perform the method described in the first aspect or any possible implementation thereof. Exemplarily, the computing device cluster may include one or more computing devices.

[0028] Sixthly, this application provides a computer program product containing instructions that, when executed by a cluster of computing devices, cause the cluster of computing devices to perform the method described in the first aspect or any possible implementation thereof. Exemplarily, the cluster of computing devices may include one or more computing devices.

[0029] It is understood that the beneficial effects of the second to sixth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of a wireless cellular network provided in an embodiment of this application;

[0031] Figure 2 This is a schematic diagram of the architecture of a wireless cell parameter optimization system provided in an embodiment of this application;

[0032] Figure 3 This is a schematic diagram of the structure of a cell parameter configuration model provided in an embodiment of this application;

[0033] Figure 4 This is a schematic diagram of the architecture of another wireless cell parameter optimization system provided in an embodiment of this application;

[0034] Figure 5 This is a schematic diagram of another cell parameter configuration model provided in an embodiment of this application;

[0035] Figure 6 This is a flowchart illustrating a cell parameter configuration method provided in an embodiment of this application;

[0036] Figure 7 This is a schematic diagram illustrating the steps of constructing a cell parameter configuration model according to an embodiment of this application;

[0037] Figure 8 This is a schematic diagram of the structure of a cell parameter configuration device provided in an embodiment of this application;

[0038] Figure 9 This is a schematic diagram of the structure of a computing device provided in an embodiment of this application;

[0039] Figure 10 This is a schematic diagram of the structure of a computing device cluster provided in an embodiment of this application;

[0040] Figure 11 This is a schematic diagram of another computing device cluster structure provided in an embodiment of this application;

[0041] Figure 12 This is a schematic diagram of another cell parameter configuration device provided in an embodiment of this application. Detailed Implementation

[0042] In this article, the term "and / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. The symbol " / " in this article indicates that the related objects are in an "or" relationship; for example, A / B means A or B.

[0043] The terms "first" and "second," etc., used in the specification and claims herein are used to distinguish different objects, not to describe a specific order of objects. For example, "first response message" and "second response message," etc., are used to distinguish different response messages, not to describe a specific order of response messages.

[0044] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0045] In the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more, for example, multiple processing units means two or more processing units, multiple elements means two or more elements, etc.

[0046] First, some of the technical terms used in this application will be introduced.

[0047] (1) Wireless cellular network

[0048] A wireless cellular network refers to a wireless access network, which is an intermediate device connecting mobile user terminals (such as mobile phones) and the core network. It generally includes various wireless base stations and the network they form.

[0049] (2) Wireless base station

[0050] A wireless base station refers to a set of signal communication equipment installed at a specific geographical location within a wireless cellular network. It provides mobile communication services to the area its signal covers, enabling mobile users to access the mobile network and conduct activities such as calls and data transmission. In this embodiment, the wireless base station can also be referred to as a "base station." For example, as... Figure 1 As shown, this wireless cellular network contains three base stations. Four mobile terminals are located within the signal coverage area of ​​these three base stations.

[0051] (3) Wireless Cell

[0052] A wireless cell refers to an area covered by signals emitted by a base station or a portion of its antennas in a mobile communication system. Within this area, mobile communication devices (such as mobile phones) can communicate with the wireless cell via wireless signals. In this embodiment, the wireless cell can also be referred to as a "cell".

[0053] (4) Match point

[0054] A site refers to a wireless cellular network within a specific geographical area belonging to the same operator. For example, in Beijing, a wireless cellular network built by operator A can be considered one site, while a wireless cellular network built by operator B can be another site.

[0055] The technical solution provided in this application will be described below.

[0056] For example, Figure 2 This illustration shows a schematic diagram of the architecture of a wireless cell parameter optimization system provided in an embodiment of this application. Figure 2As shown, the wireless cell parameter optimization system 200 may include: at least one base station 210, a model library 220, and a control center 230 for the wireless cellular network. The base station 210 is primarily used to provide mobile communication services to its signal coverage area. For example, the signal coverage area of ​​the base station 210 may be divided into multiple cells. When there are multiple base stations 210, these multiple base stations 210 belong to the same operator. The base station 210 may upload traffic statistics data (hereinafter referred to as "call statistics data") of each cell in its signal coverage area, and / or parameter configuration data of each cell, to the control center 230 in real time or periodically (e.g., every 1 second, 1 minute, or 1 hour). For example, the call statistics data may include: cell status data and / or measured values ​​of cell performance indicators. The cell status data may include, but is not limited to, one or more of the following: the average number of users per unit time period, the average number of active users per unit time period, the average length of data packets per unit time period, and the proportion of low channel quality indicator (CQI) reports per unit time period. Cell performance metrics may include, but are not limited to, one or more of the following: average downlink perceived rate per user per unit time, average uplink perceived rate per user per unit time, and average latency of data packets transmitted by users per unit time. For example, cell performance metrics can be obtained through pre-definition. Cell parameter configuration data may include, but is not limited to, one or more of the following: cell antenna transmit power and tilt angle, initial modulation and coding scheme (MCS) value, and MCS adjustment step size. The parameter configuration data may include the configuration values ​​of the parameters to be optimized in the cell parameters. If cell parameters are not configured, the values ​​of the cell parameters can be default initial values, etc., without limitation here.

[0057] The model library 220 is mainly used to store the first feature extraction module 221, the traffic scenario segmentation module 222, the performance evaluation modules for each parameter group (i.e., parameter group performance evaluation modules 2231 to 223j), and the global performance mixing module 224. The parameter groups are obtained by grouping the parameters to be optimized in the wireless cellular network. The parameters to be optimized in the wireless cellular network can be split into multiple decoupled parameter groups; for example, traffic-related parameters can be grouped together, allowing for joint modeling of multiple related parameters. For example, each parameter group can include at least one parameter to be optimized. The model library 220 can be configured on a cloud server or in the control center 230, depending on the actual situation; no specific limitation is made here.

[0058] The first feature extraction module 221 is mainly used to extract features from the call statistics data of each cell in the wireless cellular network to obtain the cell representation of each cell. For example, the first feature extraction module 221 can extract features from the call statistics data of cell A to obtain the cell representation of cell A. This first feature extraction module 221 can be a neural network model, which can be obtained through pre-training. When training the first feature extraction module 221, it can be pre-trained based on historical call statistics data collected from one or more locations using a self-supervised learning method. The self-supervised learning method can be as follows: for call statistics data collected at the same time in a cell, some information is masked, and other information is used to predict the masked information; while for call statistics data collected at different times in a cell, some information at certain times is masked, and information from other times is used to predict the information at the masked time.

[0059] The traffic scenario segmentation module 222 is mainly used to calculate the scenario classification information of each cell based on the cell representation of each cell, so as to obtain the probability that each cell belongs to each of the at least one traffic scenario, thereby completing the traffic scenario classification of each cell. For example, the traffic scenario segmentation module 222 can calculate the scenario classification information of cell A based on the cell representation of cell A. The traffic scenario of a cell can refer to the pattern and characteristics of user communication activities within a specific time and space in that cell, which may include, but is not limited to, one or more aspects such as user behavior, communication needs, and network usage. For example, traffic scenarios may include one or more of the following: traffic scenarios related to traffic usage (e.g., high traffic, medium traffic), traffic scenarios related to signal quality (e.g., high signal quality, low signal quality), and traffic scenarios related to signal coverage (e.g., large, medium, small coverage). The traffic scenario segmentation module 222 can calculate the probability that a cell belongs to each of the at least one traffic scenario. The number of traffic scenarios can be, but is not limited to, preset by the user. For example, the scene classification information of a certain cell can be an M-dimensional vector, where M is the number of call scenarios. Each element in this M-dimensional vector is a value between [0,1], and the sum of all elements is 1. Furthermore, the value of each element in this M-dimensional vector represents the probability that the cell belongs to a particular call scenario. For example, if the value of M is 2, and the scene classification information of the cell is (1 / 3, 2 / 3), then the probability that the cell belongs to one call scenario is 1 / 3, and the probability that it belongs to another call scenario is 2 / 3. This call scenario segmentation module 222 can be a neural network model, which can also be obtained through pre-training.

[0060] The performance evaluation module 223j for parameter group j is mainly used to calculate the performance contribution of parameter group j in a cell based on the configuration values ​​of each parameter to be optimized in the searched parameter group j and the scene classification information of a certain cell output by the traffic scene division module 222. The performance evaluation modules for each parameter group can be neural network models, or they can be obtained through pre-training.

[0061] The global performance mixing module 224 is mainly used to aggregate the performance contribution ratios of each parameter group to obtain the total performance of the cell. Specifically, the global performance mixing module 224 can first calculate the weights of the performance contribution ratios of each parameter group based on the cell's call statistics data; then, using the calculated weights, it performs a weighted calculation on the performance contribution ratios of these parameter groups to obtain the total performance of the cell. In this embodiment, the global performance mixing module 224 may include: a second feature extraction module 2241, a feature mapping module 2242, and a performance calculation module 2243. The second feature extraction module 2241 can be used to extract features from the cell's call statistics data. The feature mapping module 2242 can be used to map the features extracted by the second feature extraction module 2241 into weights representing the performance contribution ratios of each parameter group. The performance calculation module 2243 can use the weights mapped by the second feature extraction module 2241 to perform a weighted calculation on the performance contribution ratios of these parameter groups to obtain the total performance of the cell. The global performance mixing module 224 can be a neural network model, which can also be obtained through pre-training.

[0062] It should be understood that after training the first feature extraction module 221, it can be combined with the traffic scenario segmentation module 222, the performance evaluation modules for each parameter group (i.e., parameter group performance evaluation modules 2241 to 224j), and the global performance mixing module 226. The combined model is then trained based on historical call statistics data collected from one or more local stations, the parameter configuration values ​​of cells within those stations, and the measured values ​​of cell performance indicators. Here, the historical call statistics data and parameter configuration values ​​of cells can be samples, and the measured values ​​of cell performance indicators can be labels. For example, for data related to cell A, during training, the measured values ​​of cell A's performance indicators can be labels, and the global performance calculated using cell A's historical call statistics data and parameter configuration values ​​can be predictions. During training, the combined model can be trained by minimizing the loss based on the labels and predictions. After training is complete, the trained modules can be split and stored in the model library 220. Furthermore, since different locations may be divided into different parameter groups, and when the parameter groups are different, the feature mapping module 2242 needs to map out different numbers of weight values. Therefore, in the model library 220, the feature mapping module 2242 can be, but is not limited to, a module in an initialization state, which can map the features extracted by the second feature extraction module 2241 to the currently required number of weight values. For example, if the number of parameter groups is 2, the feature mapping module 2242 can map out two weight values; if the number of parameter groups is 3, the feature mapping module 2242 can map out three weight values, and one weight value is associated with one parameter group.

[0063] The above is an introduction to model library 220. Next, we will introduce control center 230.

[0064] The control center 230 can be managed by the operator to which the wireless cellular network belongs, and can configure parameters for each cell in the wireless cellular network. For example, the control center 230 can configure one or more of the following: the transmit power, tilt angle, initial MCS value, and MCS adjustment step size of the cell antenna. In this embodiment, the control center 230 may include: a call statistics data acquisition module 231, a parameter search module 232, and a parameter optimal configuration filtering module 233. Among them, the call statistics data acquisition module 231 is mainly used to collect call statistics data of each cell in the wireless cellular network. Since the call statistics data contains a lot of data, and not all of the data is related to cell performance indicators, the data collected by the call statistics data acquisition module 231 can be preprocessed. During preprocessing, the divergence of the call statistics data of a certain cell in various dimensions (such as signal coverage, number of users, data traffic, connection type, network speed, or user behavior) can be calculated using mathematical formulas, and the correlation between the call statistics data of the cell and the cell performance indicators can be calculated using mathematical formulas in each dimension. The divergence of call statistics data in a certain dimension characterizes the variability of the call statistics data in that dimension. The higher the divergence, the greater the variability, indicating that the call statistics data can provide more information in that dimension. The correlation between call statistics data in a certain dimension and cell performance indicators characterizes the impact of call statistics data on cell performance indicators in that dimension. Then, based on the divergence of the cell's call statistics data in each dimension and its correlation with cell performance indicators, the score of the cell's call statistics data in each dimension is determined. For example, the higher the correlation and the greater the divergence, the higher the score. Finally, dimensions that meet the score requirements can be selected, and the data in these dimensions can be used as the data required for parameter configuration. In this way, it can be ensured that the input data used for prediction is strongly correlated with the output and has a large variability. The parameter search module 232 is mainly used to search for the configuration values ​​of each parameter to be optimized in the preset parameter space. The optimal parameter configuration filtering module 233 is mainly used to filter the optimal configuration values ​​of each parameter to be optimized in the cell from the configuration values ​​of each parameter to be optimized searched by the parameter search module 233, based on the total performance of the cell calculated by the global performance mixing module 224.

[0065] In this embodiment, when configuring cell parameters for a new site (i.e., a new wireless cellular network), the control center 230 can obtain the required neural network module from the model library 220, and assemble the obtained neural network module with its call statistics data acquisition module 231, parameter search module 232, and optimal parameter configuration filtering module 233 to construct a cell parameter configuration model. The cell parameter configuration model may include: a call statistics data acquisition module 231, a first feature extraction module 221, a call scenario segmentation module 222, a parameter search module 232, a performance evaluation module for at least one parameter group, a global performance mixing module 224, and an optimal parameter configuration filtering module 233. For example, the cell parameter configuration model constructed by the control center 230 can be as follows: Figure 3 As shown. In Figure 3 In the process, the cell parameter configuration model 300 may include: a call statistics data acquisition module 231, a first feature extraction module 221, a call scenario division module 222, a parameter search module 232, a performance evaluation module for parameter groups 1 to k, a global performance hybrid module 224, and a parameter optimal configuration screening module 233.

[0066] After constructing the cell parameter configuration model for the new location, control center 230 can use this model to configure the parameters of the cells in the new location. The following describes the cell parameter configuration process using cell i as an example. (Continue reading...) Figure 3 The parameter configuration process for cell i may include: In S31, the control center 230 collects the call statistics data S of cell i through the call statistics data acquisition module 231. i In S32, the control center 230 communicates statistical data S through the feature extraction module 221. i Perform feature extraction to obtain the cell representation H of cell i. i In S33, the control center 230 uses the call scenario segmentation module 222 to characterize the cell H. i After processing, the scene classification information C of cell i is obtained. i In S34, the control center 230 iteratively searches for the configuration values ​​of each parameter to be optimized in parameter groups 1 to k through the parameter search module 232. Additionally, during each iteration, the control center 230 can use the call statistics data S... i Scene classification information C i Based on the configuration values ​​of each parameter to be optimized found in this iteration, the global performance contributed by each parameter group in this iteration is calculated. Specifically, in each iteration, for any parameter group, the control center 230 can evaluate the scene classification information C through the parameter group's performance evaluation module. i The configuration values ​​of each parameter to be optimized in this parameter group are processed to obtain the performance contribution of this parameter group in this iteration, which is based on the scene classification information C.i The performance contribution of each parameter group in this iteration is calculated based on the configuration values ​​of each parameter to be optimized within the parameter group found in this round of search. Simultaneously, during each iteration, the results output by the performance evaluation module of each parameter group can be input into the global performance mixing module 224. Afterwards, the control center 230 can access the statistical data S through this global performance mixing module 224. i The process involves calculating the weight of the performance contribution of each parameter group, and then aggregating the results output by the performance evaluation modules of each parameter group in this round using the calculated weights to obtain the global performance contributed by each parameter group in this iteration. For example, during this iteration, the global performance mixing module 224 can feed back its calculated global performance to the parameter search module 232, so that the parameter search module 232 can search for the parameter values ​​required for the next iteration based on the global performance calculated during this iteration. This allows the parameter search module 232 to gradually search for parameters in the direction of optimal performance, improving the parameter search effect. Furthermore, during the first iteration, the parameter search module 232 can, but is not limited to, randomly search for parameters. It should be understood that in S34, this step can end when the number of iterations reaches a preset number or the iteration duration reaches a preset duration. In S35, the control center 230 can compare the global performance calculated in each iteration by the global performance mixing module 224 with the global performance output by the parameter optimal configuration filtering module 233, filter out the optimal global performance, and select the configuration value of the parameter to be optimized related to the optimal global performance as the optimal configuration value. The optimal configuration value is then sent to cell i so that cell i can operate with the optimal configuration value. For example, assuming there are 2 iterations and 2 parameters to be optimized, the value of parameter 1 to be optimized found in the first iteration is a1, the value of parameter 2 to be optimized is a2, and the global performance calculated in the first iteration is R1. The value of parameter 1 to be optimized found in the second iteration is b1, the value of parameter 2 to be optimized is b2, and the global performance calculated in the second iteration is R2. If R1 < R2, then (b1, b2) can be selected as the optimal configuration parameter value of the parameter to be optimized.

[0067] After completing the parameter configuration for the cells in the new site, the control center 230 can periodically optimize the parameters of the cells included in that site. During optimization, the control center can also use... Figure 3The parameter configuration process shown selects the optimal configuration parameter values ​​for each cell and configures the parameters in each cell. Furthermore, when the number of parameters to be optimized changes, the control center 230 can also modify the performance evaluation modules of the parameter groups in the assembled cell parameter configuration model 300. For example, it can delete the performance evaluation module of a certain parameter group or add the performance evaluation module of another parameter group from the model library 220, so as to configure the parameters to be optimized after the change in quantity.

[0068] As can be seen from the above content. Figure 3 In the cell parameter configuration model 300 shown, the performance evaluation modules for different parameter groups are a shared and unified traffic scenario segmentation module. While calculating the probability of a cell belonging to at least one traffic scenario through a single traffic scenario segmentation module can achieve traffic scenario segmentation, its granularity is coarse and difficult to refine to specific parameters, resulting in relatively low accuracy of the selected parameters. To achieve finer-grained segmentation of traffic scenarios, this embodiment can configure a separate traffic scenario segmentation module for each parameter group. In this case, at least one traffic scenario can be segmented under each parameter group, meaning multiple traffic scenarios can be segmented for each parameter group. The traffic scenario segmentation module configured for a specific parameter group c can be used to calculate the cell's scenario classification information based on the cell's cell representation, thereby obtaining the probability that the cell belongs to each traffic scenario segmented under that parameter group c, thus completing the traffic scenario classification of the cell. In some embodiments, the scenario classification information calculated by the traffic scenario segmentation modules of each parameter group can be understood as the aforementioned... Figure 3 The scenario classification information described herein includes sub-scenario classification information. This sub-scenario classification information can also be a vector, and the dimensions of different sub-scenario classification information can be the same or different, and can be, but is not limited to, specified by the user. For example, sub-scenario classification information A can be an S-dimensional vector, and sub-scenario classification information B can be a Q-dimensional vector. Here, S can be the number of traffic scenarios divided under parameter group A, and Q can be the number of traffic scenarios divided under parameter group B. S and Q may be equal or unequal. Furthermore, each element in the S-dimensional vector is a value between [0,1], and the sum of all elements is 1. The value of an element in the S-dimensional vector represents the probability that the first cell belongs to a traffic scenario. Similarly, each element in the Q-dimensional vector is also a value between [0,1], and the sum of all elements is 1. The value of an element in the Q-dimensional vector represents the probability that the first cell belongs to a traffic scenario.

[0069] Furthermore, the aforementioned Figure 2 The wireless cell parameter optimization system 200 described herein can be modified as follows: Figure 4 The wireless cell parameter optimization system 400 shown is shown. Figure 4 The wireless cell parameter optimization system 400 shown is... Figure 2 The main difference between the wireless cell parameter optimization system 200 shown is that the traffic scenario division modules stored in their model libraries are different. Figure 4 The model library 240 in the wireless cell parameter optimization system 400 shows traffic scenario division modules for various parameter groups. This difference leads to different cell parameter configuration models assembled by the control center 230. Specifically, in... Figure 4 Under the wireless cell parameter optimization system 400 shown, the cell parameter configuration model assembled by the control center 230 can be as follows: Figure 5 As shown. Among them, Figure 5 The cell parameter configuration model 500 shown is... Figure 3 The main difference between the cell parameter configuration model 300 shown is: Figure 3 The performance evaluation modules for different parameter groups in the cell parameter configuration model 300 shown share a unified traffic scenario division module, while... Figure 5 In the cell parameter configuration model 500 shown, each parameter group's performance evaluation module has its own traffic scenario division module. It should be noted that the control center 230 utilizes... Figure 5 The process of configuring cell parameters using the cell parameter configuration model 500 shown can be found in [reference needed]. Figure 3 The relevant descriptions in the documentation will not be repeated here. Furthermore, for the training process of each module in model library 240, please refer to the aforementioned introduction to the training process of the relevant modules in model library 220; these will not be repeated here.

[0070] Furthermore, in this embodiment, regardless of Figure 2 The wireless cell parameter optimization system 200 shown is still... Figure 4 The wireless cell parameter optimization system 400 shown allows for user-specified or default parameters to be optimized at any site; no specific limitation is made here. When assembling the cell parameter configuration model, if the model library 220 / 240 does not contain a performance evaluation module for a certain parameter group, the control center 230 can initialize a performance evaluation module for that parameter group, add it to the cell parameter configuration model, train the assembled model, and then use the trained model to configure parameters for cells in new sites. The training data can include, but is not limited to, pre-collected historical call statistics data, parameter configuration values, and measured values ​​of cell performance indicators. The training process is detailed above in the description of the training process for the relevant modules in the model library 220 and will not be repeated here.

[0071] In addition, after the control center 230 constructs the cell parameter configuration model, in order to improve the adaptability of the constructed cell parameter configuration model with the new site, it can first collect some call statistics data from the new site, then use this data to fine-tune the cell parameter configuration model, and finally use the fine-tuned cell parameter configuration model to configure the parameters of the new site.

[0072] When fine-tuning the cell parameter configuration model, to improve the training and fine-tuning effect, data with significant differences can be collected from new sites to increase data diversity. Then, this highly differentiated data can be used for fine-tuning. The data from new sites can be collected using a strategy based on the similarity of cell traffic scenarios. Specifically, for each parameter f to be optimized, a similarity graph can be constructed. Each node in the similarity graph represents a cell; two connected nodes indicate that the two cells are similar, and the weight of the edge between nodes represents the similarity between the two cells. Assume that the parameter f has L... i Given a parameter f with several possible values, each represented by a color, data sampling under that parameter f can be transformed into a graph coloring problem. Specifically, nodes at both ends of the same edge (similar cells) should be assigned different colors (different parameter values) as much as possible to encourage similar cells to explore different parameter configurations, thereby increasing the diversity of data collection and improving the model's learning performance. Assume an edge e between nodes p and q... i,pq The weight on is w i,pq The parameter values ​​of the two end nodes are a and a, respectively. i,p and a i,q If edge e is defined i,pq The loss on is cost(w) i,pq ,a i,p ,a i,q If the global loss is L = ∑cost(w), then the global loss is: L = ∑cost(w) i,pq ,a i,p ,a i,q To minimize the global loss L, we need to define an edge as a conflicting edge if both ends have the same color. Specifically, to minimize the number of conflicting edges, for cost(w)... i,pq ,a i,p ,a i,q It can be defined that when a i,p =a i,q At that time, cost(w) i,pq ,a i,p ,a i,q )=1, when a i,p ≠a i,q At that time, cost(w) i,pq ,a i,p ,a i,q) = 0. When we want to minimize the weight of conflicting edges, for cost(w) i,pq ,a i,p ,a i,q It can be defined that when a i,p =a i,q At that time, cost(w) i,pq ,a i,p ,a i,q ) = w i,pq , when a i,p ≠a i,q At that time, cost(w) i,pq ,a i,p ,a i,q = 0. The similarity between cells can be calculated using the scene partitioning information of different cells for parameter f. For example, if the scene partitioning information of cells p and q for parameter f are c... i,p and c i,q c i,p and c i,q If both are vectors of length N with element values ​​between [0,1], and the sum of all elements in each vector is 1, then the similarity between these two neighborhoods can be, but is not limited to, as follows:

[0073]

[0074] In other words, when fine-tuning the model, different parameter values ​​can be configured for cells with similar call traffic scenarios to collect diverse data, thereby improving sampling efficiency and the effectiveness of model fine-tuning. For cells with dissimilar call traffic scenarios, either the same or different parameter values ​​can be configured. For example, the data needed to calculate the call traffic scenario similarity of a cell can be collected under default parameter values. It should be understood that configuring cell parameters before model fine-tuning is only to facilitate the collection of diverse data; subsequent call statistics data need to be collected from the cell again after model fine-tuning is completed, and then used... Figure 3 or Figure 5 The cell configuration model shown configures the parameters of the cell.

[0075] The above is an introduction to the wireless cell parameter optimization system provided in the embodiments of this application. In the above-described wireless cell parameter optimization system, since pre-trained neural network modules are stored in the model library, the cell parameter configuration model can be customized and assembled based on the settings of new locations, making model deployment more flexible and eliminating the need for re-pre-training for each new location, thus reducing deployment costs. Simultaneously, since the neural network modules included in the assembled cell parameter configuration model are pre-trained or initialized, recommendations can be made directly in new locations, or only a small amount of data is needed for model fine-tuning, greatly shortening the optimization cycle for new locations. Furthermore, the assembled cell parameter configuration model can also classify cells into different traffic scenarios using real call statistics data within the cell, enabling differentiated parameter recommendations for cells with different traffic scenarios and improving parameter optimization effectiveness.

[0076] It should be noted that the above Figure 3 or Figure 5 The modules in the cell parameter configuration model shown can be added or removed according to actual needs, and the modified scheme is still within the protection scope of this application.

[0077] Next, based on the above, a cell parameter configuration method provided by an embodiment of this application will be introduced. It should be understood that the cell parameter configuration here includes both initial configuration and subsequent optimization.

[0078] For example, Figure 6 This diagram illustrates a flowchart of a cell parameter configuration method provided in an embodiment of this application. It is understood that this method can be executed by any device, equipment, platform, or device cluster with computing and processing capabilities. For example, this method can be executed by a cell parameter configuration device, which can be implemented in software and / or hardware and can be configured in a computing device, typically a server. Figure 6 As shown, the cell parameter configuration method may include the following steps:

[0079] S601. Obtain the first call statistics data of the first cell in the target site. The target site is a wireless cellular network belonging to the same operator in a certain area.

[0080] In this embodiment, when configuring parameters for cells in the target site, call statistics data of the cells in the target site can be collected first. Specifically, when configuring parameters for the first time, the cells in the target site can be run with default parameters to collect call statistics data. When configuring parameters for subsequent times, historical call statistics data collected before this configuration can be used. The specific approach depends on the actual situation and is not limited here. For example, the first call statistics data may include: the status data of the first cell, such as: the average number of users per unit time period, the average number of active users per unit time period, the average length of data packets per unit time period, etc.

[0081] S602. Based on the first call statistics data, classify the call scenarios of the first cell to obtain the scenario classification information of the first cell. The scenario classification information is used to indicate the probability that the first cell belongs to each call scenario in at least one call scenario.

[0082] In this embodiment, after obtaining the first call statistics data of the first cell in the target site, the call scenarios of the first cell can be classified based on the first call statistics data to obtain the scenario classification information of the first cell. The scenario classification information of the first cell indicates the probability that the first cell belongs to each of at least one call scenario. Thus, the call scenarios of the first cell can be classified using the actual call statistics data in the first cell. For example, feature extraction can be performed on the first call statistics data first, for instance, using the aforementioned first feature extraction module 221 to obtain the cell representation of the first cell. Then, the scenario classification information of the first cell is calculated based on the cell representation. For example, the cell representation of the first cell can be processed using the aforementioned call scenario division module 222 or the parameter scenario division module of each parameter group to calculate the scenario classification information of the first cell.

[0083] S603. Based on the first call statistics data and scene classification information, determine the optimal configuration value of the parameters to be optimized in the first cell.

[0084] In this embodiment, after obtaining the scene classification information and the first call statistics data of the first cell, the optimal configuration values ​​of the parameters to be optimized in the first cell can be determined at least using these two sets of data. For the process of determining the optimal configuration values ​​of the parameters to be optimized in the first cell, please refer to the foregoing. Figure 3 The parameter configuration process for cell i is not detailed here.

[0085] In this way, when configuring parameters in a cell at a central office location, the cell is divided into traffic scenarios based on real call statistics data within the cell. The configuration of parameters to be optimized in the cell is determined by the traffic scenarios obtained from the division. This allows for differentiated parameter recommendations for cells with different traffic scenarios, thereby improving the parameter optimization effect.

[0086] In some embodiments, prior to S602, a cell configuration model can be constructed to configure parameters of the first cell. Specifically, such as... Figure 7 As shown, the process of constructing a cell configuration model may include: in S701, based on the configuration requirements of the target site, such as the parameters that need to be configured, selecting neural network modules related to the parameter configuration of the first cell from a model library containing multiple neural network modules (e.g., the aforementioned model library 220 or 240). The neural network modules in the model library are pre-trained or initialized. For example, the selected neural network modules may include: the aforementioned first feature extraction module 221, the traffic scenario segmentation module 222, or the parameter scenario segmentation modules for each parameter group (e.g., the aforementioned...). Figure 5 The parameter scenario partitioning module 2221 of parameter group 1, etc.), and the performance evaluation modules of each parameter group (such as those mentioned above) Figure 3 Or the performance evaluation module of parameter group 1 in 5, etc.), global performance hybrid module 226, etc. In S702, at least the selected neural network modules related to the parameter configuration of the first cell are assembled to obtain a cell parameter configuration model. After selecting the neural network modules, these modules can be assembled together to obtain a cell parameter configuration model. The cell parameter configuration model is used to process at least the first call statistics data to obtain the optimal configuration value of the parameters to be optimized in the first cell. For example, the assembled cell parameter configuration model can be, but is not limited to, the following: Figure 3 The cell parameter configuration model 300 shown is... Figure 5 The cell parameter configuration model shown is 500. After assembling the cell parameter configuration model, it can be used to process at least the first call statistics data to obtain the optimal configuration values ​​of the parameters to be optimized in the first cell. Since the neural network modules in the model library are pre-trained or partially initialized, the cell parameter configuration model can be customized based on the configuration requirements of each site, making model deployment more flexible and eliminating the need for re-pre-training for each site, thus reducing deployment costs. Furthermore, since the modules in the assembled cell parameter configuration model are pre-trained or partially initialized, recommendations can be made directly in new sites, or only a small amount of data is needed for model fine-tuning, greatly shortening the optimization cycle for new sites.

[0087] Furthermore, after assembling the cell parameter configuration model, fine-tuning can be performed to improve its adaptability to the target site and thus enhance parameter configuration accuracy. To increase the diversity of data used for fine-tuning, differentiated parameter configuration values ​​can be assigned to cells with similar call traffic scenarios within the target site; that is, different parameter configuration values ​​are used for cells with similar call traffic scenarios. This ensures that differentiated data is collected from cells with similar call traffic scenarios. Then, the data required for fine-tuning is collected from the target site.

[0088] It is understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. In addition, the various embodiments described above can be combined according to actual conditions, and the combined solutions are still within the protection scope of this application.

[0089] Based on the methods in the above embodiments, this application also provides a cell parameter configuration device.

[0090] For example, Figure 8 A schematic diagram of a cell parameter configuration device provided in an embodiment of this application is shown. Figure 8 As shown, the cell parameter configuration device 800 includes an acquisition module 810 and a processing module 820. The acquisition module is used to acquire first call statistics data of a first cell in a target location, where the target location is a wireless cellular network belonging to the same operator within a geographical area. The processing module 820 is used to classify the first cell into call scenarios based on the first call statistics data to obtain scenario classification information for the first cell. The scenario classification information indicates the probability that the first cell belongs to each of at least one call scenario. Furthermore, based on the first call statistics data and the scenario classification information, the optimal configuration value of the parameters to be optimized in the first cell is determined.

[0091] In some embodiments, there are multiple parameters to be optimized, which are divided into multiple parameter groups. The call scenario is divided by treating multiple parameter groups as a whole, and the number of scenario classification information is 1.

[0092] In some embodiments, the scene classification information is an M-dimensional vector, where M is the number of call scenarios, each element in the M-dimensional vector is a value between [0,1], and the sum of all elements is 1. The value of an element in the M-dimensional vector represents the probability that the first cell belongs to a call scenario.

[0093] In some embodiments, there are multiple parameters to be optimized, and these multiple parameters are divided into multiple parameter groups, with each parameter group having at least one call scenario.

[0094] The scenario classification information includes: multiple sub-scenario classification information. The number of sub-scenario classification information is the same as the number of parameter groups. A sub-scenario classification information is used to indicate the probability that the first cell belongs to each traffic scenario divided under a parameter group.

[0095] In some embodiments, the multiple parameter groups include: a first parameter group and a second parameter group; the scenario classification information includes: a first sub-scenario classification information and a second sub-scenario classification information; the first sub-scenario classification information is an S-dimensional vector, where S is the number of traffic scenarios divided under the first parameter group; the second sub-scenario classification information is a Q-dimensional vector, where Q is the number of traffic scenarios divided under the second parameter group; S and Q may be equal or unequal; wherein, each element in the S-dimensional vector is a value between [0,1], and the sum of all elements is 1; the value of one element in the S-dimensional vector represents the probability that the first cell belongs to a traffic scenario; each element in the Q-dimensional vector is a value between [0,1], and the sum of all elements is 1; the value of one element in the Q-dimensional vector represents the probability that the first cell belongs to a traffic scenario.

[0096] In some embodiments, there are multiple parameters to be optimized, and these multiple parameters are divided into multiple parameter groups. At this time, when the processing module 820 performs traffic scenario classification on the first cell based on the first call statistics data to obtain the scenario classification information of the first cell, it specifically performs the following: feature extraction on the first call statistics data to obtain the cell representation of the first cell; and calculates the scenario classification information based on the cell representation of the first cell.

[0097] In some embodiments, when the processing module 820 determines the optimal configuration value of the parameter to be optimized in the first cell based on the first call statistics data and scene classification information, it is specifically used for: iteratively searching for the configuration value of the parameter to be optimized; and, in each iteration, calculating the global performance jointly contributed by each parameter group in the current iteration based on the first call statistics data, scene classification information, and the configuration value of the parameter to be optimized searched in the current iteration; and selecting the optimal configuration value from the configuration values ​​of the parameter to be optimized searched in the iterative search based on the global performance jointly contributed by each parameter group calculated in each iteration, wherein the global performance jointly contributed by each parameter group is optimal under the optimal configuration value.

[0098] In some embodiments, when the processing module 820 calculates the global performance contributed by each parameter group in the current iteration based on the first call statistics data, scene classification information, and the configuration values ​​of the parameters to be optimized found in the current iteration, it specifically performs the following: calculating the performance contribution ratio of each parameter group based on the configuration values ​​of each parameter to be optimized in each parameter group found in the current iteration and the scene classification information; calculating the weight of the performance contribution ratio of each parameter group based on the first call statistics data; and calculating the global performance contributed by each parameter group in the current iteration based on the performance contribution ratio of each parameter group and the corresponding weight.

[0099] In some embodiments, before classifying the traffic scenarios of the first cell based on the first call statistics data, the processing module 820 is further configured to: based on the configuration requirements of the target site, select neural network modules related to the parameter configuration of the first cell from a model library containing multiple neural network modules, wherein the neural network modules in the model library are pre-trained or some of them are initialized; assemble the selected neural network modules related to the parameter configuration of the first cell to obtain a cell parameter configuration model, wherein the cell parameter configuration model is used to process at least the first call statistics data to obtain the optimal configuration value of the parameters to be optimized in the first cell.

[0100] In some embodiments, the processing module 820 is further configured to: fine-tune the cell parameter configuration model, wherein the data required for fine-tuning is: collected from the target site after assigning differentiated parameter configuration values ​​to cells with similar call scenarios in the target site.

[0101] In some embodiments, the scenario classification information is a vector of length M, where M is the number of preset traffic scenarios, each element in the vector of length M is a value between [0,1], and the sum of all elements is 1. The value of an element in the vector of length M represents the probability that the traffic scenario of the first cell belongs to a preset traffic scenario.

[0102] In some embodiments, Figure 8 Both the acquisition module 810 and the processing module 820 shown can be implemented in software or in hardware. For example, the implementation of the acquisition module 810 will be described below. Similarly, the implementation of the processing module 820 can refer to the implementation of the acquisition module 810.

[0103] As an example of a software functional unit, module 810 may include code running on a computing instance. The computing instance may include at least one of a physical host (computing device), a virtual machine, or a container. Further, the aforementioned computing instance may be one or more. For example, module 810 may include code running on multiple hosts / virtual machines / containers. It should be noted that the multiple hosts / virtual machines / containers used to run the code may be distributed within the same region or in different regions. Further, the multiple hosts / virtual machines / containers used to run the code may be distributed within the same availability zone (AZ) or in different AZs, each AZ including one or more geographically proximate data centers. Typically, a region may include multiple AZs.

[0104] Similarly, multiple hosts / virtual machines / containers used to run this code can be distributed within the same Virtual Private Cloud (VPC) or across multiple VPCs. Typically, a VPC is set up within a region. Communication between two VPCs within the same region, as well as between VPCs in different regions, requires a communication gateway to be set up within each VPC to enable interconnection between VPCs.

[0105] As an example of a hardware functional unit, the acquisition module 810 may include at least one computing device, such as a server. Alternatively, the acquisition module 810 may also be a device implemented using an application-specific integrated circuit (ASIC) or a programmable logic device (PLD). The PLD may be implemented using a complex programmable logical device (CPLD), a field-programmable gate array (FPGA), generic array logic (GAL), or any combination thereof.

[0106] The multiple computing devices included in the acquisition module 810 can be distributed in the same region or in different regions. Similarly, the multiple computing devices included in the acquisition module 810 can be distributed in the same Availability Zone (AZ) or in different AZs. Likewise, the multiple computing devices included in the acquisition module 810 can be distributed in the same Virtual Private Cloud (VPC) or in multiple VPCs. These multiple computing devices can be any combination of computing devices such as servers, ASICs, PLDs, CPLDs, FPGAs, and GALs.

[0107] It should be noted that, in other embodiments, the acquisition module 810 can be used to execute any step in the cell parameter configuration method described in the above embodiments, and the processing module 820 can also be used to execute any step in the cell parameter configuration method described in the above embodiments. Furthermore, the acquisition module 810 can also be combined with the processing module 820 to be responsible for executing any step in the cell parameter configuration method described in the above embodiments. In addition, the steps implemented by the acquisition module 810 and the processing module 820 can also be specified as needed, and different steps in the cell parameter configuration method described in the above embodiments can be implemented by the acquisition module 810 and the processing module 820 respectively. Figure 8 The diagram shows all the functions of the community parameter configuration device 800.

[0108] This application also provides a computing device 900. For example... Figure 9 As shown, the computing device 900 includes a bus 902, a processor 904, a memory 906, and a communication interface 908. The processor 904, the memory 906, and the communication interface 908 communicate with each other via the bus 902. The computing device 900 can be a server or a terminal device. It should be understood that this application does not limit the number of processors and memories in the computing device 900.

[0109] The 902 bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 9 The bus 904 is represented by a single line, but this does not mean that there is only one bus or one type of bus. The bus 904 may include a path for transmitting information between various components of the computing device 900 (e.g., memory 906, processor 904, communication interface 908).

[0110] Processor 904 may include any one or more processors such as a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), or a digital signal processor (DSP).

[0111] The memory 906 may include volatile memory, such as random access memory (RAM). The processor 904 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid state drive (SSD).

[0112] The memory 906 stores executable program code, and the processor 904 executes the executable program code to implement the aforementioned functions respectively. Figure 8 The functions of the acquisition module 810 and processing module 820 shown are used to implement the cell parameter configuration method described in the above embodiments. That is, the memory 906 stores instructions for executing the cell parameter configuration method described in the above embodiments.

[0113] Alternatively, the memory 906 stores executable code, and the processor 904 executes the executable code to implement the aforementioned functions respectively. Figure 8 The cell parameter configuration device 800 shown in the figure functions to implement the cell parameter configuration method described in the above embodiments. That is, the memory 906 stores instructions for executing the cell parameter configuration method described in the above embodiments.

[0114] The communication interface 908 uses transceiver modules, such as, but not limited to, network interface cards and transceivers, to enable communication between the computing device 900 and other devices or communication networks.

[0115] This application also provides a computing device cluster. The computing device cluster includes at least one computing device. The computing device can be a server, such as a central server, an edge server, or a local server in a local data center. In some embodiments, the computing device can also be a terminal device such as a desktop computer, a laptop computer, or a smartphone.

[0116] like Figure 10As shown, the computing device cluster includes at least one computing device 900. The memory 906 of one or more computing devices 900 in the computing device cluster may store the same instructions for executing the cell parameter configuration method described in the above embodiments.

[0117] In some possible implementations, the memory 906 of one or more computing devices 900 in the computing device cluster may also store partial instructions for executing the cell parameter configuration method described in the above embodiments. In other words, a combination of one or more computing devices 900 can jointly execute the instructions for executing the cell parameter configuration method described in the above embodiments.

[0118] It should be noted that the memory 906 in different computing devices 900 within the computing device cluster can store different instructions, each used to execute the aforementioned instructions. Figure 8 The illustrated cell parameter configuration device 800 performs some of its functions. Specifically, the instructions stored in the memory 906 of different computing devices 900 can implement the functions of one or more modules in the acquisition module 810 and processing module 820.

[0119] In some possible implementations, one or more computing devices in a computing device cluster can be connected via a network. This network can be a wide area network (WAN) or a local area network (LAN), etc. Figure 11 One possible implementation is shown. For example... Figure 11 As shown, two computing devices 900A and 900B are connected via a network. Specifically, they are connected to the network through communication interfaces in each computing device. In this possible implementation, the memory 906 in computing device 900A stores instructions for executing the functions of the acquisition module 810. Simultaneously, the memory 906 in computing device 900B stores instructions for executing the functions of the processing module 820.

[0120] It should be understood that Figure 11 The functions of the computing device 900A shown can also be performed by multiple computing devices 900. Similarly, the functions of the computing device 900B can also be performed by multiple computing devices 900.

[0121] This application also provides another computing device cluster. The connection relationships between the computing devices in this computing device cluster can be similarly referred to... Figure 10 and Figure 11 The connection method of the computing device cluster is different in that the memory 906 of one or more computing devices 900 in the computing device cluster can store the same instructions for executing the methods in the above embodiments.

[0122] In some possible implementations, the memory 906 of one or more computing devices 900 in the computing device cluster may also store partial instructions for executing the aforementioned cell parameter configuration method. In other words, a combination of one or more computing devices 900 can jointly execute the instructions for executing the aforementioned cell parameter configuration method.

[0123] Based on the methods described in the above embodiments, this application also provides a cell parameter configuration device. Please refer to... Figure 12 , Figure 12 This is a schematic diagram of another cell parameter configuration device provided in an embodiment of this application. Figure 12 As shown, the cell parameter configuration device 1200 includes one or more processors 1201 and interface circuits 1202. Optionally, the cell parameter configuration device 1200 may also include a bus 1203. Wherein:

[0124] Processor 1201 can be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed through integrated logic circuits in the hardware of processor 1201 or through software instructions. Processor 1201 can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods and steps disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor.

[0125] The interface circuit 1202 can be used to send or receive data, instructions or information. The processor 1201 can use the data, instructions or other information received by the interface circuit 1202 to process the data, instructions or other information, and can send the processed information out through the interface circuit 1202.

[0126] Optionally, the model inference performance enhancement device 1200 also includes a memory, which may include read-only memory and random access memory, and provides operation instructions and data to the processor. A portion of the memory may also include non-volatile random access memory (NVRAM).

[0127] Optionally, the memory stores executable software modules or data structures, and the processor can execute corresponding operations by calling the operation instructions stored in the memory (which may be stored in the operating system).

[0128] Optionally, the interface circuit 1202 can be used to output the execution results of the processor 1201.

[0129] It should be noted that the functions of the processor 1201 and the interface circuit 1202 can be implemented through hardware design, software design, or a combination of hardware and software; no restrictions are imposed here.

[0130] It should be understood that each step of the above method embodiments can be accomplished by hardware logic circuits or software instructions in a processor.

[0131] Based on the methods in the above embodiments, this application provides a computer-readable storage medium including computer program instructions. When executed by a cluster of computing devices including at least one computing device, the computer program instructions cause the cluster of computing devices to perform the methods in the above embodiments. Exemplarily, the computer-readable storage medium can be any available medium that the computing device can store, or a data storage device such as a data center containing one or more available media. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives).

[0132] Based on the methods in the above embodiments, this application provides a computer program product containing instructions that, when executed by a cluster of computing devices containing at least one computing device, cause the cluster of computing devices to perform the methods in the above embodiments.

[0133] It is understood that the processor in the embodiments of this application can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. A general-purpose processor can be a microprocessor or any conventional processor.

[0134] The method steps in the embodiments of this application can be implemented in hardware or by a processor executing software instructions. The software instructions can consist of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, portable hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can reside in an ASIC.

[0135] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0136] It is understood that the various numerical designations used in the embodiments of this application are merely for descriptive convenience and are not intended to limit the scope of the embodiments of this application.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the embodiments of this application.

Claims

1. A method for configuring cell parameters, characterized in that, The method includes: Obtain the first call statistics data of the first cell in the target site, wherein the target site is a wireless cellular network belonging to the same operator in a geographical area; Based on the first call statistics data, the first cell is classified into call scenarios to obtain the scenario classification information of the first cell. The scenario classification information is used to indicate the probability that the first cell belongs to each call scenario in at least one call scenario. Based on the first call statistics data and the scenario classification information, the optimal configuration value of the parameters to be optimized in the first cell is determined.

2. The method according to claim 1, characterized in that, The parameters to be optimized are multiple, and the multiple parameters to be optimized are divided into multiple parameter groups. The call scenario is obtained by treating the multiple parameter groups as a whole. The number of scenario classification information is 1.

3. The method according to claim 2, characterized in that, The scenario classification information is an M-dimensional vector, where M is the number of the call scenarios. Each element in the M-dimensional vector is a value between [0,1], and the sum of all elements is 1. The value of an element in the M-dimensional vector represents the probability that the first cell belongs to one of the call scenarios.

4. The method according to claim 1, characterized in that, There are multiple parameters to be optimized, and these multiple parameters to be optimized are divided into multiple parameter groups. Each parameter group is divided into at least one call scenario. The scenario classification information includes: multiple sub-scenario classification information, the number of which is the same as the number of parameter groups, and one sub-scenario classification information is used to indicate the probability that the first cell belongs to each traffic scenario divided under a parameter group.

5. The method according to claim 4, characterized in that, The multiple parameter groups include: a first parameter group and a second parameter group. The scene classification information includes: a first sub-scene classification information and a second sub-scene classification information. The first sub-scene classification information is an S-dimensional vector, where S is the number of traffic scenes divided under the first parameter group. The second sub-scene classification information is a Q-dimensional vector, where Q is the number of traffic scenes divided under the second parameter group. S and Q may be equal or unequal. In this S-dimensional vector, each element is a value between [0,1], and the sum of all elements is 1. The value of an element in the S-dimensional vector represents the probability that the first cell belongs to a traffic scenario. Each element in the Q-dimensional vector is a value between [0,1], and the sum of all elements is 1. The value of an element in the Q-dimensional vector represents the probability that the first cell belongs to a traffic scenario.

6. The method according to any one of claims 1-5, characterized in that, The step of classifying the call scenarios of the first cell based on the first call statistics data to obtain scenario classification information for the first cell includes: Feature extraction is performed on the first call statistics data to obtain the cell representation of the first cell; Based on the cell representation of the first cell, the scene classification information is calculated.

7. The method according to any one of claims 1-6, characterized in that, There are multiple parameters to be optimized, and these multiple parameters to be optimized are divided into multiple parameter groups; The step of determining the optimal configuration value of the parameters to be optimized in the first cell based on the first call statistics data and the scene classification information includes: The configuration values ​​of the parameters to be optimized are searched iteratively, and in each iteration, the global performance contributed by each parameter group in the current iteration is calculated based on the first call statistics data, the scene classification information and the configuration values ​​of the parameters to be optimized found in the current iteration. Based on the global performance contributed by each parameter group in each iteration, the optimal configuration value is selected from the configuration values ​​of the parameters to be optimized obtained through iterative search, wherein the global performance contributed by each parameter group is optimal under the optimal configuration value.

8. The method according to claim 7, characterized in that, The calculation of the global performance contributed by each parameter group in this iteration, based on the first call statistics data, the scene classification information, and the configuration values ​​of the parameters to be optimized found in this iteration, includes: Based on the configuration values ​​of each parameter to be optimized in each parameter group searched in this iteration and the scene classification information, the performance contribution of each parameter group is calculated. Based on the first set of statistics data, calculate the weight of the performance contribution of each parameter group; Based on the performance contribution ratio and corresponding weight of each parameter group, the global performance contributed by each parameter group in this iteration is calculated.

9. The method according to any one of claims 1-8, characterized in that, Before classifying the call traffic scenarios of the first cell based on the first call statistics data, the process also includes: Based on the configuration requirements of the target site, neural network modules related to the parameter configuration of the first cell are selected from a model library containing multiple neural network modules, wherein the neural network modules in the model library are obtained through pre-training or initialization; At least the selected neural network modules related to the parameter configuration of the first cell are assembled to obtain a cell parameter configuration model, wherein the cell parameter configuration model is used to process at least the first call statistics data to obtain the optimal configuration value of the parameter to be optimized in the first cell.

10. The method according to claim 9, characterized in that, Also includes: The cell parameter configuration model is fine-tuned. The data required for fine-tuning is collected from the target site after assigning differentiated parameter configuration values ​​to cells with similar call scenarios in the target site.

11. A cell parameter configuration device, characterized in that, include: The acquisition module is used to acquire the first call statistics data of the first cell in the target site, wherein the target site is a wireless cellular network belonging to the same operator in a geographical area; The processing module is used to classify the first cell into traffic scenarios based on the first call statistics data to obtain the scenario classification information of the first cell. The scenario classification information is used to indicate the probability that the first cell belongs to each of the at least one traffic scenarios. The processing module is further configured to determine the optimal configuration value of the parameters to be optimized in the first cell based on the first call statistics data and the scene classification information.

12. The apparatus according to claim 11, characterized in that, The parameters to be optimized are multiple, and the multiple parameters to be optimized are divided into multiple parameter groups. The call scenario is obtained by treating the multiple parameter groups as a whole. The number of scenario classification information is 1.

13. The apparatus according to claim 12, characterized in that, The scenario classification information is an M-dimensional vector, where M is the number of the call scenarios. Each element in the M-dimensional vector is a value between [0,1], and the sum of all elements is 1. The value of an element in the M-dimensional vector represents the probability that the first cell belongs to one of the call scenarios.

14. The apparatus according to claim 1, characterized in that, The parameters to be optimized are multiple, and the multiple parameters to be optimized are divided into multiple parameter groups. Each parameter group is divided into at least one of the call scenarios. The scenario classification information includes multiple sub-scenario classification information, the number of which is the same as the number of parameter groups. Each sub-scenario classification information is used to indicate the probability that the first cell belongs to each traffic scenario divided under a parameter group.

15. The apparatus according to claim 14, characterized in that, The multiple parameter groups include: a first parameter group and a second parameter group. The scene classification information includes: a first sub-scene classification information and a second sub-scene classification information. The first sub-scene classification information is an S-dimensional vector, where S is the number of traffic scenes divided under the first parameter group. The second sub-scene classification information is a Q-dimensional vector, where Q is the number of traffic scenes divided under the second parameter group. S and Q may be equal or unequal. In this S-dimensional vector, each element is a value between [0,1], and the sum of all elements is 1. The value of an element in the S-dimensional vector represents the probability that the first cell belongs to a traffic scenario. Each element in the Q-dimensional vector is a value between [0,1], and the sum of all elements is 1. The value of an element in the Q-dimensional vector represents the probability that the first cell belongs to a traffic scenario.

16. The apparatus according to any one of claims 10-15, characterized in that, When the processing module performs traffic scenario classification on the first cell based on the first call statistics data to obtain scenario classification information for the first cell, it is specifically used for: Feature extraction is performed on the first call statistics data to obtain the cell representation of the first cell; Based on the cell representation of the first cell, the scene classification information is calculated.

17. The apparatus according to any one of claims 10-16, characterized in that, There are multiple parameters to be optimized, and these multiple parameters to be optimized are divided into multiple parameter groups; When determining the optimal configuration value of the parameters to be optimized in the first cell based on the first call statistics data and the scene classification information, the processing module is specifically used for: The configuration values ​​of the parameters to be optimized are searched iteratively, and in each iteration, the global performance contributed by each parameter group in the current iteration is calculated based on the first call statistics data, the scene classification information and the configuration values ​​of the parameters to be optimized found in the current iteration. Based on the global performance contributed by each parameter group calculated in each iteration, the optimal configuration value is selected from the configuration values ​​of the parameters to be optimized found in the iterative search, wherein the global performance contributed by each parameter group is optimal under the optimal configuration value.

18. The apparatus according to claim 17, characterized in that, When the processing module calculates the global performance contributed by each parameter group in the current iteration based on the first call statistics data, the scene classification information, and the configuration values ​​of the parameters to be optimized found in the current iteration, it is specifically used for: Based on the configuration values ​​of each parameter to be optimized in each parameter group searched in this iteration and the scene classification information, the performance contribution of each parameter group is calculated. Based on the first set of statistics data, calculate the weight of the performance contribution of each parameter group; Based on the performance contribution ratio and corresponding weight of each parameter group, the global performance contributed by each parameter group in this iteration is calculated.

19. The apparatus according to any one of claims 10-18, characterized in that, Before classifying the call traffic scenarios of the first cell based on the first call statistics data, the processing module is further used for: Based on the configuration requirements of the target site, neural network modules related to the parameter configuration of the first cell are selected from a model library containing multiple neural network modules, wherein the neural network modules in the model library are obtained through pre-training or initialization; At least the selected neural network modules related to the parameter configuration of the first cell are assembled to obtain a cell parameter configuration model, wherein the cell parameter configuration model is used to process at least the first call statistics data to obtain the optimal configuration value of the parameter to be optimized in the first cell.

20. The apparatus according to claim 19, characterized in that, The processing module is further configured to: The cell parameter configuration model is fine-tuned. The data required for fine-tuning is collected from the target site after assigning differentiated parameter configuration values ​​to cells with similar call scenarios in the target site.

21. A cell parameter configuration device, characterized in that, Includes at least one processor and interface; The at least one processor obtains program instructions through the interface; The at least one processor is configured to execute the program line instructions to implement the method as described in any one of claims 1-10.

22. A computing device cluster, characterized in that, It includes at least one computing device, each computing device including a processor and memory; The processor of the at least one computing device is configured to execute instructions stored in the memory of the at least one computing device to cause the cluster of computing devices to perform the method as described in any one of claims 1-10.

23. A computer-readable storage medium, characterized in that, The method includes computer program instructions that, when executed by a cluster of computing devices, cause the cluster of computing devices to perform the method as described in any one of claims 1-10, wherein the cluster of computing devices includes at least one computing device.

24. A computer program product containing instructions, characterized in that, When the instruction is executed by the computing device cluster, the computing device cluster causes the computing device cluster to perform the method as described in any one of claims 1-10, wherein the computing device cluster includes at least one computing device.

Images