A wireless network optimization method, server, medium and system

Abstract

The application provides a wireless network optimization method, server, medium and system. The method comprises the following steps: acquiring a wireless network problem, and acquiring a preconfigured Bayesian network model corresponding to the wireless network problem; acquiring current influence factor item values corresponding to influence factor item included in a directed acyclic graph (DAG) of the preconfigured Bayesian network model from a network optimization database; and analyzing the current influence factor item values by using the preconfigured Bayesian network model, and acquiring an optimization measure corresponding to the wireless network problem. The method solves the problem that an existing network optimization model cannot accurately acquire an optimization scheme for solving a network problem.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a wireless network optimization method, server, medium, and system. Background Technology

[0002] Currently, with the increasing scale of 5G network construction, the requirements for wireless network optimization are gradually increasing. Existing wireless network analysis and optimization methods mainly utilize intelligent algorithms, including conventional shallow learning methods and deep learning methods, to mine and analyze relevant network data, establish network optimization models, and solve wireless network problems to optimize wireless networks.

[0003] However, existing network optimization models are mainly built through black-box analysis of network data. They utilize a small number of high-quality solutions and the model's parameters are not sufficiently learned. As a result, existing network optimization models often fail to accurately obtain optimization solutions for network problems. Summary of the Invention

[0004] This application provides a wireless network optimization method, server, medium, and system to address the problem that existing network optimization models cannot accurately obtain optimization solutions for network problems.

[0005] In a first aspect, this application provides a wireless network optimization method, comprising: acquiring a wireless network problem and acquiring a pre-configured Bayesian network model corresponding to the wireless network problem; based on the directed acyclic graph (DAG) of the pre-configured Bayesian network model, acquiring the current influencing factor values ​​corresponding to the influencing factor items included in the DAG from a network optimization database; and using the pre-configured Bayesian network model to analyze the current influencing factor values ​​and acquire optimization measures corresponding to the wireless network problem.

[0006] In one specific implementation, the method for configuring the pre-configured Bayesian network model includes: obtaining a directed acyclic graph (DAG) corresponding to a preset wireless network problem, wherein the DAG includes influencing factor terms and optimization measure terms; obtaining training data from a daily network optimization library and a large network optimization database based on the preset wireless network problem, the influencing factor terms, and the optimization measure terms; training a Bayesian network model using the training data and the DAG, and detecting the goodness of fit of the Bayesian network model using the Bayesian information criterion; and loading the Bayesian network model when the goodness of fit is determined to be less than a preset threshold value, so as to obtain the optimization measures corresponding to the preset wireless network problem using the Bayesian network model.

[0007] In one specific implementation, the step of obtaining training data from a daily network optimization library and a network optimization database based on the preset wireless network problem, the influencing factors, and the optimization measures includes: querying network problem events corresponding to the preset wireless network problem and the optimization measures from the daily network optimization library; when the network problem event is determined to be an improved event, obtaining the processing object, processing time, and processing measures of the network problem event; obtaining the influencing factor values ​​corresponding to the processing object within a preset time period before the processing time from the network optimization database, and merging the influencing factor values ​​and the processing measures to form training data.

[0008] In one specific implementation, training a Bayesian network model using the training data and the DAG includes: dividing the training data into target domain data and source domain data according to the type of processor; generating artificial data using the target domain data and mixing the artificial data with the target domain data to obtain augmented target domain data; performing dimensionality increase processing on the source domain data and the augmented target domain data respectively; classifying the dimensionality-increased source domain data and the augmented target domain data using a Naive Bayes classifier to obtain probability values ​​for each dimension-increased source domain data and augmented target domain data; obtaining Bayesian network parameters corresponding to each node in the DAG based on the probability values ​​and probability parameters of each dimension-increased source domain data and augmented target domain data, and the DAG; and forming a Bayesian network model based on the DAG and the Bayesian network parameters corresponding to each node in the DAG.

[0009] In one specific implementation, the step of detecting the goodness of fit of the Bayesian network model using the Bayesian information criterion includes: using the Bayesian information criterion, the Bayesian network parameters corresponding to each node in the DAG, and the source domain data and the augmented target domain data after the dimension increase processing, to detect the goodness of fit of the DAG and obtain the goodness of fit of the DAG.

[0010] Secondly, this application provides a server, comprising: a model deployment module, configured to acquire a wireless network problem and acquire a pre-configured Bayesian network model corresponding to the wireless network problem; a sample generation module, configured to acquire, based on the directed acyclic graph (DAG) of the pre-configured Bayesian network model, the current influencing factor values ​​corresponding to the influencing factor items included in the DAG from a network optimization database; the model deployment module is further configured to use the pre-configured Bayesian network model to analyze the current influencing factor values ​​and acquire optimization measures corresponding to the wireless network problem.

[0011] In one specific embodiment, the server further includes: an acquisition module, configured to acquire a directed acyclic graph (DAG) corresponding to a preset wireless network problem, the DAG including influencing factor terms and optimization measure terms; a sample generation module, further configured to acquire training data from a daily network optimization library and a network optimization database based on the preset wireless network problem, the influencing factor terms, and the optimization measure terms; a model training module, configured to train a Bayesian network model using the training data and the DAG; a model detection module, configured to detect the goodness of fit of the Bayesian network model using the Bayesian information criterion; and a model deployment module, further configured to load the Bayesian network model when the goodness of fit is determined to be less than a preset threshold, so as to acquire the optimization measures corresponding to the preset wireless network problem using the Bayesian network model.

[0012] In one specific implementation, the sample generation module is specifically used to: query the network problem events corresponding to the preset wireless network problems and the optimization measures from the daily network optimization database; when the network problem event is determined to be an improved event, obtain the processing object, processing time, and processing measures of the network problem event; obtain the influencing factor values ​​corresponding to the processing object within a preset time before the processing time from the network optimization database, and merge the influencing factor values ​​and the processing measures to form training data.

[0013] In one specific implementation, the model training module is specifically used for: dividing the training data into target domain data and source domain data according to the type of the processor; generating artificial data using the target domain data, and mixing the artificial data with the target domain data to obtain augmented target domain data; performing dimensionality increase processing on the source domain data and the augmented target domain data respectively; classifying the dimensionality-increased source domain data and the augmented target domain data using a Naive Bayes classifier to obtain probability values ​​for each dimension-increased source domain data and augmented target domain data; obtaining Bayesian network parameters corresponding to each node in the DAG based on the probability values ​​and probability parameters of each dimension-increased source domain data and augmented target domain data, and the DAG; and forming a Bayesian network model based on the DAG and the Bayesian network parameters corresponding to each node in the DAG.

[0014] In one specific implementation, the model detection module is specifically used to: detect the DAG using the Bayesian information criterion, the Bayesian network parameters corresponding to each node in the DAG, and the source domain data and augmented target domain data after the dimension increase processing, and obtain the goodness of fit of the DAG.

[0015] Thirdly, this application provides a server, including: a processor, a memory, and a communication interface; the memory is used to store executable instructions of the processor; wherein the processor is configured to execute the wireless network optimization method described in the first aspect by executing the executable instructions.

[0016] Fourthly, this application provides a readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the wireless network optimization method described in the first aspect.

[0017] Fifthly, this application provides a wireless network optimization system, including: a daily network optimization library, a network optimization database, and a server as described in the second to fourth aspects.

[0018] This application provides a wireless network optimization method, server, medium, and system. The method includes: identifying a wireless network problem and obtaining a pre-configured Bayesian network model corresponding to the problem; based on the directed acyclic graph (DAG) of the pre-configured Bayesian network model, obtaining the current values ​​of influencing factors included in the DAG from a network optimization database; and using the pre-configured Bayesian network model to analyze the current values ​​of the influencing factors and obtain optimization measures corresponding to the wireless network problem. Compared to existing network optimization models, which cannot accurately obtain optimization solutions for network problems, this application obtains a pre-configured Bayesian network model corresponding to the wireless network problem, obtains the current values ​​of influencing factors based on the DAG of the Bayesian network model, and analyzes the current values ​​of the influencing factors using the Bayesian network model to obtain optimization measures corresponding to the wireless network problem. Therefore, by using the pre-configured Bayesian network model, corresponding optimization measures can be accurately obtained for wireless network problems, solving the problem that existing network optimization models cannot accurately obtain optimization solutions for network problems. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A wireless network optimization system architecture diagram is provided in this application;

[0021] Figure 2 A flowchart illustrating an embodiment of a wireless network optimization method provided in this application;

[0022] Figure 3 This is the directed acyclic graph corresponding to the wireless network problem "VoLTE high backflow 2G cell";

[0023] Figure 4 A flowchart illustrating a second embodiment of a wireless network optimization method provided in this application;

[0024] Figure 5 A schematic diagram illustrating the loading method of the Bayesian network model provided in this application;

[0025] Figure 6 This application provides a schematic diagram of the structure of a server embodiment.

[0026] Figure 7 A schematic diagram of another server embodiment provided in this application;

[0027] Figure 8 This is a schematic diagram of another server embodiment provided in this application. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments made by those skilled in the art under the guidance of these embodiments are within the scope of protection of this application.

[0029] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0030] First, let me explain the terms used in this application:

[0031] A Bayesian network model, also known as a confidence network, consists of a Directed Acyclic Graph (DAG) and a Conditional Probability Table (CPT). Constructing a Bayesian network model involves two aspects: first, building the model structure, i.e., the DAG; and second, obtaining the conditional probability distributions at each node of the network. Node variables can be abstractions of any problem, such as test values, observed phenomena, or opinions solicited. It is suitable for expressing and analyzing uncertain and probabilistic events, applied to decision-making that conditionally depends on multiple control factors, and can make inferences from incomplete, imprecise, or uncertain knowledge or information.

[0032] A Directed Acyclic Graph (DAG) consists of nodes representing variables and directed edges connecting these nodes. Nodes represent random variables, and directed edges represent the relationships between nodes (from a parent node to its child node). The strength of these relationships is expressed using conditional probabilities, while those without parent nodes use prior probabilities. A DAG depicts the dependencies between random variables, which can be direct or indirect. Two random variables directly connected by an arrow have a direct relationship, while those with an intermediate variable have an indirect relationship.

[0033] Currently, with the increasing scale of 5G network construction, the requirements for wireless network optimization are gradually increasing. Existing wireless network analysis and optimization methods mainly utilize intelligent algorithms, including conventional shallow learning methods and deep learning methods, to mine and analyze relevant network data, establish network optimization models, and solve wireless network problems to optimize wireless networks.

[0034] However, existing network optimization models are mainly built through black-box analysis of network data. They utilize a small number of high-quality solutions and the model's parameters are not sufficiently learned. As a result, existing network optimization models often fail to accurately obtain optimization solutions for network problems.

[0035] Based on the above-mentioned technical problems, the technical conception process of this application is as follows: How to solve the problem that existing network optimization models cannot accurately obtain optimization solutions for solving network problems.

[0036] The wireless network optimization scheme of this application will be described in detail below.

[0037] Figure 1 This application provides a diagram of a wireless network optimization system architecture. Figure 1As shown, the wireless network optimization system includes: a daily network optimization library 11, a network optimization big data library 12, and a server. The server includes a model deployment module 101, a sample generation module 102, an acquisition module 103, a model training module 104, and a model detection module 105.

[0038] The model deployment module 101 acquires the wireless network problem and the corresponding pre-configured Bayesian network model. The sample generation module 102, based on the directed acyclic graph (DAG) of the pre-configured Bayesian network model, retrieves the current values ​​of the influencing factors included in the DAG from the network optimization database 12. The model deployment module 101 uses the pre-configured Bayesian network model to analyze the current influencing factor values ​​and acquire the optimization measures corresponding to the wireless network problem.

[0039] The pre-configured Bayesian network model is also configured by the server. The acquisition module 103 acquires the directed acyclic graph (DAG) corresponding to the preset wireless network problem. The DAG includes influencing factor terms and optimization measure terms. The sample generation module 102 acquires training data from the daily network optimization library 11 and the network optimization big data library 12 based on the preset wireless network problem, influencing factor terms, and optimization measure terms. The model training module 104 trains the Bayesian network model using the training data and the DAG. The model detection module 105 uses the Bayesian information criterion to detect the goodness of fit of the Bayesian network model. The model deployment module 101 loads the Bayesian network model when the goodness of fit is determined to be less than a preset threshold, so as to use the Bayesian network model to acquire the optimization measures corresponding to the preset wireless network problem.

[0040] The technical solution of this application will now be described in detail through specific embodiments. It should be noted that the following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0041] Figure 2 This is a flowchart illustrating an embodiment of a wireless network optimization method provided in this application. See also... Figure 2 The wireless network optimization method specifically includes the following steps:

[0042] Step S201: Obtain the wireless network problem and obtain the pre-configured Bayesian network model corresponding to the wireless network problem.

[0043] In this embodiment, a wireless network problem is obtained; for example, the wireless network problem can be a "VoLTE high backflow 2G cell". A pre-configured Bayesian network model corresponding to the wireless network problem is obtained. This Bayesian network model includes a directed acyclic graph (DAG) and Bayesian network parameters corresponding to each node in the DAG.

[0044] Step S202: Based on the pre-configured Bayesian network model's directed acyclic graph (DAG), obtain the current influencing factor values ​​corresponding to the influencing factor items included in the DAG from the network optimization database.

[0045] In this embodiment, the pre-configured Bayesian network model's directed acyclic graph (DAG) includes influencing factor terms and optimization measures terms.

[0046] Figure 3 This is a directed acyclic graph corresponding to the wireless network problem "VoLTE high backflow 2G cell". For example... Figure 3 As shown, the directed acyclic graph (DAG) includes influencing factors such as "cell scenario", "cell frequency band", "surrounding 4G cell density", "soft switch not turned on", "no 4G coverage", "weak 4G coverage" and "high-quality but poor 4G coverage", as well as optimization measures such as "building a new 4G station", "entering the weak coverage DAG", "entering the poor quality DAG" and "guiding the opening of the soft switch".

[0047] Based on the directed acyclic graph (DAG), the current values ​​of the influencing factors included in the DAG are obtained from the network optimization database. For example, the current value of the influencing factor "surrounding 4G cell density" is obtained, which is the current surrounding 4G cell density.

[0048] Step S203: Using the pre-configured Bayesian network model, analyze the current influencing factor values ​​to obtain the optimization measures corresponding to the wireless network problem.

[0049] In this embodiment, a pre-configured Bayesian network model is used to analyze the values ​​of current influencing factors obtained from the network optimization database, thereby obtaining optimization measures corresponding to the wireless network problem. For example, using the pre-configured Bayesian network model to analyze the current influencing factor value "current surrounding 4G cell density", the optimization measure corresponding to the wireless network problem "VoLTE high backflow 2G cell" is "build a new 4G site".

[0050] In this embodiment, a wireless network problem is identified, and a pre-configured Bayesian network model corresponding to the problem is obtained. Based on the directed acyclic graph (DAG) of the pre-configured Bayesian network model, the current values ​​of the influencing factors included in the DAG are obtained from a network optimization database. The pre-configured Bayesian network model is then used to analyze these current influencing factor values ​​to obtain optimization measures corresponding to the wireless network problem. Compared to existing network optimization models, which cannot accurately obtain optimization solutions for network problems, this application obtains a pre-configured Bayesian network model corresponding to the wireless network problem, acquires the current influencing factor values ​​based on the DAG of the Bayesian network model, and analyzes these values ​​using the Bayesian network model to obtain optimization measures corresponding to the wireless network problem. Therefore, by using a pre-configured Bayesian network model, corresponding optimization measures can be accurately obtained for wireless network problems, solving the problem that existing network optimization models cannot accurately obtain optimization solutions for network problems.

[0051] Figure 4 This is a flowchart illustrating a second embodiment of a wireless network optimization method provided in this application. Figure 2 Based on the illustrated embodiment, see also Figure 4 The configuration method for this pre-configured Bayesian network model specifically includes the following steps:

[0052] Step S401: Obtain the directed acyclic graph (DAG) corresponding to the preset wireless network problem. The DAG includes influencing factor items and optimization measure items.

[0053] For example, the preset wireless network problem can be "VoLTE high backflow 2G cell". Obtain the directed acyclic graph (DAG) corresponding to the preset wireless network problem. This DAG is constructed based on expert experience and includes influencing factor items and optimization measure items.

[0054] like Figure 3 As shown, the influencing factors can be "cell scenario", "cell frequency band", "surrounding 4G cell density", "soft switch not turned on", "no 4G coverage", "weak 4G coverage", and "high-quality but poor-quality 4G". The optimization measures can be "building a new 4G base station", "entering a weak coverage DAG", "entering a poor-quality DAG", and "guiding the activation of the soft switch".

[0055] Step S402: Based on the preset wireless network problem, the influencing factor item, and the optimization measure item, obtain training data from the daily network optimization library and the network optimization big data library.

[0056] In this embodiment, the daily network optimization database stores specific optimization solutions and their implementation effects for problem cells and sites in daily wireless network optimization and special optimization work. These solutions are categorized into expert solutions and solutions provided by ordinary optimization personnel, and are marked according to the personnel who provided the solutions. The network optimization database stores performance indicators, parameters, measurement reports, and engineering parameters related to the wireless network.

[0057] Specifically, the system queries the daily network optimization database for the network problem and the network problem event corresponding to the optimization measure. When the network problem event is determined to be an improved event, the system obtains the processing object, processing time, and processing measure for the network problem event. The system retrieves the value of the influencing factor item corresponding to the processing object within a preset time before the processing time from the network optimization database, and merges the value of the influencing factor item and the processing measure to form training data.

[0058] In this embodiment, the daily network optimization database records the processing object, processing time, processing measures, and processing effect of each network problem. The database is used to query preset wireless network problems and the corresponding network problem events for each optimization measure. When a network problem event is determined to be an improved event, the processing object, processing time, and processing measures for that network problem event are obtained.

[0059] For example, if the preset wireless network problem is "VoLTE high backhaul 2G cell" and the optimization measure is "build a new 4G site", then the system will query the daily network optimization database for network problem events with the wireless network problem "VoLTE high backhaul 2G cell" and the handling measure "build a new 4G site". When the handling effect of this network problem event is determined to be "improved", the system will obtain the handling object, handling time, and handling measure for this network problem event. For example, the handling object of this network problem event can be the cell involved in the network problem event.

[0060] In this embodiment, the values ​​of influencing factors corresponding to the processing object within a preset time period before the processing time are obtained from the network optimization database. These influencing factor values ​​are then combined with the processing measure to form training data. For example, the values ​​of influencing factors corresponding to the cell involved in the network problem event within a preset time period before the processing time are obtained from the network optimization database; for instance, the surrounding 4G cell density value of the cell one day before the processing time is obtained. This surrounding 4G cell density value is then combined with the processing measure "building a new 4G base station" to form training data.

[0061] Step S403: Using the training data and the DAG, train a Bayesian network model, and use the Bayesian information criterion to detect the goodness of fit of the Bayesian network model.

[0062] In this embodiment, a Bayesian network model is trained using training data and a Directed Acyclic Graph (DAG). Specifically, the training data is divided into target domain data and source domain data according to the type of personnel processing the data. Artificial data is generated using the target domain data, and this artificial data is mixed with the target domain data to obtain augmented target domain data. The source domain data and the augmented target domain data are then subjected to dimensionality-increasing processing. A Naive Bayes classifier is used to classify the dimensionality-increasing source domain data and the augmented target domain data to obtain probability values ​​for each dimension-increasing source domain data and augmented target domain data. Based on the probability values ​​and probability parameters of each dimension-increasing source domain data and augmented target domain data, as well as the DAG, the Bayesian network parameters corresponding to each node in the DAG are obtained. A Bayesian network model is formed based on the DAG and the Bayesian network parameters corresponding to each node in the DAG.

[0063] In this embodiment, the solutions for wireless network problems stored in the daily network optimization library are categorized into expert solutions and solutions for ordinary optimization personnel based on the type of personnel handling the issues. Therefore, the training data is divided into target domain data and source domain data according to the type of personnel handling the issues. Specifically, the target domain data corresponds to expert personnel, while the source domain data corresponds to ordinary optimization personnel.

[0064] Since the number of solutions processed by experts is relatively small, and the corresponding target domain data is also relatively small, it is necessary to augment the target domain data by generating artificial data using the target domain data and then mixing the artificial data with the target domain data to obtain augmented target domain data.

[0065] Specifically, the Synthetic Minority Oversampling Technique (SMOTE algorithm) is applied to utilize the target domain data. T Generate artificial data A Data from human sources A Data with target domain T Mixing yields amplified target domain data. new .

[0066] Data from the source domain S and augmented target domain data new Perform dimensionality increase processing on the source domain data. S and augmented target domain data new Add one dimension to each to form Data' S and Data' new For example, for source domain data Data Sand augmented target domain data new Add a marker to each, representing the source domain data (Data). S Add a marker "0" to expand the target domain data. new Add marker "1" to form Data' S and Data' new .

[0067] The source domain data after dimensionality increase is processed using a Naive Bayes classifier. S and augmented target domain data. new Perform classification processing to obtain the processed source domain data for each dimension. S and augmented target domain data. new The probability value is P(σ=1|x).

[0068] Add processed source domain data (Data) according to each dimension. S and augmented target domain data. new Given the probability values, probability parameters, and the directed acyclic graph (DAG), obtain the Bayesian network parameters corresponding to each node in the DAG.

[0069] Specifically, for each dimension, the processed source domain data (Data') is added. S Apply the formula:

[0070]

[0071] Get the source domain data after processing for each dimension. S probability parameters For each dimension, add processed augmented target domain data (Data'). new Its probability parameter The value is 1.

[0072] For each node in the DAG, apply the formula:

[0073]

[0074] Obtain the probability value of "when this node and its parent node occur" for each node, which is the Bayesian network parameter for each node. Among them, D l Add processed source domain data to the dimension. S and augmented target domain data. new ; Let I be the probability parameters for Data'S and Data'new; I is the value function, which determines whether the current node Xi and its parent node appear in D. lIf the value is 1, then the value is 1; otherwise, the value is 0.

[0075] A Bayesian network model is formed based on the DAG and the Bayesian network parameters corresponding to each node in the DAG.

[0076] In this embodiment, the goodness of fit of the Bayesian network model is detected using the Bayesian information criterion. Specifically, after the Bayesian network model is formed, the goodness of fit of the DAG is detected using the Bayesian information criterion, the Bayesian network parameters corresponding to each node in the DAG, and the source domain data and target domain data after the dimension is increased.

[0077] In this embodiment, the Bayesian Information Criterion (BIC) is used to detect the goodness of fit of the Bayesian network model. Let the nodes in the DAG of the Bayesian network model be a1, a2, ..., a s Then apply the formula: BIC = log Pr(a1,a2,...,a) s The goodness-of-fit of the DAG is calculated as ) + d / 2 * log n. s ) represents the joint probability distribution of the Bayesian network model, d is twice the number of nodes in the DAG, and n is the source domain data after dimensionality enhancement. S and augmented target domain data. new The quantity.

[0078] Specifically, model detection can be achieved by calling the Bayesian network module in R as follows:

[0079] library(bnlearn) # Load the Bayesian network module

[0080] bn.score<-score(x=dag,data=train_data,type="bic")

[0081] Where dag is a directed acyclic graph (DAG), and train_data is Data' S and Data' new “bic” is the Bayesian information criterion.

[0082] Obtain the goodness-of-fit value bn.score. If bn.score is less than the preset threshold h / d, the DAG passes the test; if bn.score is greater than or equal to the preset threshold, the DAG structure is tuned.

[0083] Step S404: When the goodness of fit is determined to be less than a preset threshold, the Bayesian network model is loaded to obtain the optimization measures corresponding to the preset wireless network problem using the Bayesian network model.

[0084] In this embodiment, in order to persist the trained Bayesian network model, the model is loaded by the server, and a Representational State Transfer (REST) ​​interface or other forms of service interface are provided so that the Bayesian network model can be accessed by calling the service interface.

[0085] Specifically, the loading method of the Bayesian network model is as follows: Figure 5 As shown. Figure 5 This is a schematic diagram illustrating the loading method of the Bayesian network model provided in this application. (See diagram below.) Figure 5 As shown, the network is deployed in an integrated manner. Classifier-1, classifier-2, and classifier-3 are all Bayesian network models that output the probability values ​​of each optimization measure based on the input wireless network problem. The models are connected through an ensemble node, enabling them to access each other.

[0086] In this embodiment, a directed acyclic graph (DAG) corresponding to a preset wireless network problem is obtained. This DAG includes influencing factor terms and optimization measure terms. Training data is obtained from a daily network optimization library and a large network optimization database based on the preset wireless network problem, the influencing factor terms, and the optimization measure terms. A Bayesian network model is trained using this training data and the DAG, and the goodness of fit of the Bayesian network model is detected using the Bayesian information criterion. When the goodness of fit is determined to be less than a preset threshold, the Bayesian network model is loaded to obtain the optimization measures corresponding to the preset wireless network problem. Therefore, the Bayesian network model of this application, trained based on a directed acyclic graph (DAG) constructed from expert experience and optimization schemes for improved network problem events recorded in the daily network optimization library, and tested using the Bayesian information criterion, further improves the accuracy of the optimization schemes obtained using this model.

[0087] The following are embodiments of the apparatus described in this application, which can be used to execute the embodiments of the method described in this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the method described in this application.

[0088] Figure 6 This application provides a schematic diagram of the structure of a server embodiment; as shown below. Figure 6As shown, the server 60 includes a model deployment module 61 and a sample generation module 62. The model deployment module 61 is used to acquire a wireless network problem and obtain a pre-configured Bayesian network model corresponding to the problem. The sample generation module 62 is used to retrieve the current influencing factor values ​​corresponding to the influencing factor items included in the DAG of the pre-configured Bayesian network model from a network optimization database. The model deployment module 61 is also used to analyze the current influencing factor values ​​using the pre-configured Bayesian network model to obtain optimization measures corresponding to the wireless network problem.

[0089] The server provided in this application embodiment can execute the technical solutions shown in the above method embodiments. Its implementation principle and beneficial effects are similar, and will not be described again here.

[0090] Figure 7 A schematic diagram of another server embodiment provided in this application; as shown Figure 7 As shown, the server 70 includes: a model deployment module 61, a sample generation module 62, an acquisition module 71, a model training module 72, and a model detection module 73. The acquisition module 71 acquires a directed acyclic graph (DAG) corresponding to a preset wireless network problem, the DAG including influencing factor terms and optimization measure terms. The sample generation module 62 acquires training data from a daily network optimization library and a large network optimization database based on the preset wireless network problem, the influencing factor terms, and the optimization measure terms. The model training module 72 trains a Bayesian network model using the training data and the DAG. The model detection module 73 detects the goodness of fit of the Bayesian network model using the Bayesian information criterion. The model deployment module 61 loads the Bayesian network model when the goodness of fit is determined to be less than a preset threshold, so as to acquire the optimization measures corresponding to the preset wireless network problem using the Bayesian network model.

[0091] The server provided in this application embodiment can execute the technical solutions shown in the above method embodiments. Its implementation principle and beneficial effects are similar, and will not be described again here.

[0092] In one possible implementation, the sample generation module 62 is specifically used to query the network problem event corresponding to the preset wireless network problem and the optimization measure item from the daily network optimization library; when the network problem event is determined to be an improved event, the processing object, processing time and processing measure of the network problem event are obtained; the value of the influencing factor item corresponding to the processing object within a preset time before the processing time is obtained from the network optimization database, and the value of the influencing factor item and the processing measure are merged to form training data.

[0093] In one possible implementation, the model training module 72 is specifically used to divide the training data into target domain data and source domain data according to the type of processor; generate artificial data using the target domain data, and mix the artificial data with the target domain data to obtain augmented target domain data; perform dimensionality increase processing on the source domain data and the augmented target domain data respectively; use a Naive Bayes classifier to classify the dimensionality-increased source domain data and the augmented target domain data to obtain the probability values ​​of the dimensionality-increased source domain data and the augmented target domain data for each dimension; obtain the Bayesian network parameters corresponding to each node in the DAG based on the probability values ​​and probability parameters of the dimensionality-increased source domain data and the augmented target domain data for each dimension; and form a Bayesian network model based on the DAG and the Bayesian network parameters corresponding to each node in the DAG.

[0094] In one possible implementation, the model detection module 73 is specifically used to detect the DAG by utilizing the Bayesian information criterion, the Bayesian network parameters corresponding to each node in the DAG, and the source domain data and the augmented target domain data after the dimension is increased, so as to obtain the goodness of fit of the DAG.

[0095] The server provided in this application embodiment can execute the technical solutions shown in the above method embodiments. Its implementation principle and beneficial effects are similar, and will not be described again here.

[0096] Figure 8 This is a schematic diagram of the structure of another data server provided in this application. Figure 8 As shown, the server 80 includes: a processor 81, a memory 82, and a communication interface 83; wherein, the memory 82 is used to store executable instructions of the processor 81; the processor 81 is configured to execute the technical solutions in any of the foregoing method embodiments by executing the executable instructions.

[0097] Optionally, the memory 82 can be either standalone or integrated with the processor 81.

[0098] Optionally, when the memory 82 is a device independent of the processor 81, the server 80 may further include a bus 84 for connecting the aforementioned devices.

[0099] The server is used to execute the technical solutions in any of the aforementioned method embodiments. Its implementation principle and technical effect are similar, and will not be described again here.

[0100] This application also provides a readable storage medium storing a computer program thereon, which, when executed by a processor, implements the technical solutions provided in any of the foregoing embodiments.

[0101] This application also provides a wireless network optimization system, including a daily network optimization library, a network optimization database, and a server as provided in any of the foregoing embodiments.

[0102] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0103] 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 or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A wireless network optimization method, characterized in that, include: Obtain the wireless network problem and the corresponding pre-configured Bayesian network model; Based on the pre-configured Bayesian network model, the directed acyclic graph (DAG) is used to obtain the current values ​​of the influencing factors corresponding to the influencing factors included in the DAG from the network optimization database. Using the pre-configured Bayesian network model, the values ​​of the current influencing factors are analyzed to obtain the optimization measures corresponding to the wireless network problem; The configuration method for the pre-configured Bayesian network model includes: Obtain the directed acyclic graph (DAG) corresponding to the preset wireless network problem, wherein the DAG includes influencing factor items and optimization measure items; Based on the preset wireless network problem, the influencing factors, and the optimization measures, training data is obtained from the daily network optimization library and the network optimization database. Using the training data and the DAG, a Bayesian network model is trained, and the goodness of fit of the Bayesian network model is detected using the Bayesian information criterion. When the goodness of fit is determined to be less than a preset threshold, the Bayesian network model is loaded to obtain the optimization measures corresponding to the preset wireless network problem using the Bayesian network model.

2. The wireless network optimization method according to claim 1, characterized in that, The step of obtaining training data from the daily network optimization library and the network optimization database based on the preset wireless network problem, the influencing factors, and the optimization measures includes: Query the preset wireless network problems and the network problem events corresponding to the optimization measures from the daily network optimization database; When the network problem event is determined to be an improved event, the processing object, processing time, and processing measures of the network problem event are obtained; The values ​​of influencing factors corresponding to the processing object within a preset time period before the processing time are obtained from the network optimization database. The values ​​of influencing factors and the processing measures are then combined to form training data.

3. The wireless network optimization method according to claim 1 or 2, characterized in that, The step of training a Bayesian network model using the training data and the DAG includes: The training data is divided into target domain data and source domain data according to the type of personnel processing the data. Artificial data is generated using the target domain data, and the artificial data is mixed with the target domain data to obtain augmented target domain data; The source domain data and the augmented target domain data are respectively subjected to dimensionality increase processing; The Naive Bayes classifier is used to classify the source domain data and the augmented target domain data after dimensionality increase processing, so as to obtain the probability value of the source domain data and the augmented target domain data after each dimensionality increase processing. Based on the probability values ​​and probability parameters of the processed source domain data and the augmented target domain data for each dimension, as well as the DAG, obtain the Bayesian network parameters corresponding to each node in the DAG. A Bayesian network model is formed based on the DAG and the Bayesian network parameters corresponding to each node in the DAG.

4. The wireless network optimization method according to claim 3, characterized in that, The method of detecting the goodness of fit of the Bayesian network model using the Bayesian information criterion includes: The DAG is detected using the Bayesian information criterion, the Bayesian network parameters corresponding to each node in the DAG, and the source domain data and target domain data after the dimension increase processing, so as to obtain the goodness of fit of the DAG.

5. A server, characterized in that, include: The model deployment module is used to obtain wireless network problems and the pre-configured Bayesian network model corresponding to the wireless network problems. The sample generation module is used to obtain the current values ​​of the influencing factors corresponding to the influencing factors included in the DAG from the network optimization database based on the directed acyclic graph (DAG) of the pre-configured Bayesian network model. The model deployment module is also used to analyze the current influencing factor values ​​using the pre-configured Bayesian network model to obtain optimization measures corresponding to the wireless network problem. Also includes: The acquisition module is used to acquire the directed acyclic graph (DAG) corresponding to the preset wireless network problem. The DAG includes influencing factor items and optimization measure items. The sample generation module is also used to obtain training data from the daily network optimization library and the network optimization database based on the preset wireless network problem, the influencing factor item and the optimization measure item; The model training module is used to train a Bayesian network model using the training data and the DAG; The model detection module is used to detect the goodness of fit of the Bayesian network model using the Bayesian information criterion. The model deployment module is further configured to load the Bayesian network model when the goodness of fit is determined to be less than a preset threshold, so as to use the Bayesian network model to obtain the optimization measures corresponding to the preset wireless network problem.

6. The server according to claim 5, characterized in that, The sample generation module is specifically used for: Query the preset wireless network problems and the network problem events corresponding to the optimization measures from the daily network optimization database; When the network problem event is determined to be an improved event, the processing object, processing time, and processing measures of the network problem event are obtained; The values ​​of influencing factors corresponding to the processing object within a preset time period before the processing time are obtained from the network optimization database. The values ​​of influencing factors and the processing measures are then combined to form training data.

7. The server according to claim 5 or 6, characterized in that, The model training module is specifically used for: The training data is divided into target domain data and source domain data according to the type of personnel processing the data. Artificial data is generated using the target domain data, and the artificial data is mixed with the target domain data to obtain augmented target domain data; The source domain data and the augmented target domain data are respectively subjected to dimensionality increase processing; The Naive Bayes classifier is used to classify the source domain data and the augmented target domain data after dimensionality increase processing, so as to obtain the probability value of the source domain data and the augmented target domain data after each dimensionality increase processing. Based on the probability values ​​and probability parameters of the processed source domain data and the augmented target domain data for each dimension, as well as the DAG, obtain the Bayesian network parameters corresponding to each node in the DAG. A Bayesian network model is formed based on the DAG and the Bayesian network parameters corresponding to each node in the DAG.

8. The server according to claim 7, characterized in that, The model detection module is specifically used for: The DAG is detected using the Bayesian information criterion, the Bayesian network parameters corresponding to each node in the DAG, and the source domain data and target domain data after the dimension increase processing, so as to obtain the goodness of fit of the DAG.

9. A server, characterized in that, include: Processor, memory, communication interface; The memory is used to store the executable instructions of the processor; The processor is configured to execute the wireless network optimization method of any one of claims 1 to 4 by executing the executable instructions.

10. A readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the wireless network optimization method according to any one of claims 1 to 4.

11. A wireless network optimization system, characterized in that, include: The daily network optimization library, the network optimization database, and the server as described in any one of claims 5 to 8.

Images