Wireless network poor-quality cell root cause analysis method and device and medium

Abstract

The invention provides a wireless network poor-quality cell root cause analysis method and device and a medium. The method comprises the following steps: acquiring multi-dimensional characteristic index sample data of a plurality of wireless network cells; according to the characteristic index sample data, evaluating the plurality of classification models to screen out an optimal classification model; according to the optimal classification model, selecting the first N key feature indexes with the highest influence degree on the identified poor-quality cell in the feature index sample data; clustering all feature indexes in the feature index sample data to obtain M root cause classifications; and calculating a comprehensive score of each root cause classification according to the number of all the feature indexes and the number and importance of all the key feature indexes in the M root cause classifications of the poor-quality cell, and determining a problem primary cause and a problem secondary cause of the poor-quality cell according to the ranking of the comprehensive scores. According to the method, the device and the medium, the problem that the root cause analysis of the cell problem with poor wireless network quality generally depends on artificial experience in the prior art can be solved.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a method, device and medium for analyzing root causes of poor quality cells in wireless networks. Background Art

[0002] Poor quality cells refer to cells with one or more poor quality problems in network communication quality. The types of poor quality problems in cells involve multiple aspects, such as voice service connection rate, call drop rate, packet loss rate, data service drop rate, access success rate, user perceived rate, etc. At present, the identification of the types of problems in poor quality cells mainly adopts the traditional network optimization (abbreviated as: network optimization) method, specifically, first collect various key performance indicators (Key Performance Indication, KPI) of the cell. Then, the network optimization staff can rely on their expert knowledge and experience to set thresholds for each KPI respectively, and then judge the type of poor quality problem of the cell based on the thresholds of these KPIs and the various KPIs of the cell. Then, the network optimization staff can analyze various data, manually check various network problems such as coverage, interference, capacity, etc., and use their experience to conduct root cause analysis and formulate optimization solutions, so as to optimize the network problems.

[0003] At present, root cause analysis of poor wireless network quality cells generally relies on manual experience and various tool software for analysis, which has many problems:

[0004] 1. There is a shortage of personnel with professional skills, and too many people without expertise has become an optimization bottleneck. 2. Network problems are complex, the causes are diverse, and the problems are difficult to locate. 3. Problems are repeatedly handled, with poor timeliness and long cycles. 4. Network problem mining has certain limitations, the real pain points are not clearly located, the manpower investment cost is high, and the benefits are low. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a method, device and medium for root cause analysis of poor quality cells in wireless networks in view of the above-mentioned deficiencies in the prior art, so as to solve the problem that the prior art generally relies on manual experience in root cause analysis of poor quality cells in wireless networks.

[0006] In a first aspect, the present invention provides a method for analyzing root causes of poor quality cells in a wireless network, comprising:

[0007] Obtain multi-dimensional characteristic indicator sample data of multiple wireless network cells;

[0008] According to the characteristic indicator sample data, multiple classification models are evaluated to screen out the optimal classification model;

[0009] According to the optimal classification model, the top N key characteristic indicators with the highest impact on identifying poor-quality cells in the characteristic indicator sample data are selected, where 20≤N≤40;

[0010] Cluster all the characteristic indicators in the characteristic indicator sample data to obtain M root cause classifications, where M <N;

[0011] For each poor quality cell among the multiple wireless network cells, a comprehensive score of each root cause classification is calculated according to the number of all characteristic indicators and the number and importance of all key characteristic indicators in the M root cause classifications of the poor quality cell, and the main cause and secondary cause of the problem of the poor quality cell are determined according to the ranking of the comprehensive score.

[0012] Furthermore, the step of evaluating multiple classification models according to the characteristic index sample data to select the optimal classification model specifically includes:

[0013] Taking the multi-dimensional characteristic index sample data of each poor quality cell among the multiple wireless network cells as a positive sample, and taking the multi-dimensional characteristic index sample data of each non-poor quality cell among the multiple wireless network cells as a negative sample, to construct a classification training set;

[0014] Using the classification training set, respectively train the multiple classification models to obtain each of the classification models after training;

[0015] Calculate the final evaluation score of each classification model after training according to the accuracy of each classification model after training, the accuracy of balanced data, the area score under the receiver operating characteristic ROC curve, the F1 evaluation score and the training time;

[0016] The trained classification model with the highest final evaluation score is used as the optimal classification model.

[0017] Furthermore, the final evaluation score of each classification model after training is calculated according to the accuracy of each classification model after training, the accuracy of balanced data, the area score under the receiver operating characteristic ROC curve, the F1 evaluation score and the training time, specifically including:

[0018] The final evaluation score of each classification model after training is calculated by the following formula:

[0019] Final_Score=sum(Accuracy+Balance_Accuracy+Roc_Auc_Score

[0020] +F1_Score)-TimeTaken / 1000

[0021] Among them, Final_Score is the final evaluation score of the classification model, Accuracy is the accuracy of the classification model, Balance_Accuracy is the balanced data accuracy of the classification model, Roc_Auc_Score is the area score under the ROC curve of the classification model, F1_Score is the F1 evaluation score of the classification model, and TimeTaken is the training time of the classification model.

[0022] Furthermore, the selecting, according to the optimal classification model, the top N key characteristic indicators with the highest influence on identifying poor quality cells in the characteristic indicator sample data specifically includes:

[0023] Determining the importance of each feature indicator in the feature indicator sample data according to the training result of the optimal classification model, wherein the importance of the feature indicator is used to measure the influence of the feature indicator on the prediction target of the optimal classification model;

[0024] The importance of each characteristic indicator in the characteristic indicator sample data is arranged in descending order, and the first N characteristic indicators after the descending order are selected as the first N key characteristic indicators with the highest influence on identifying poor quality cells.

[0025] Furthermore, clustering all characteristic indicators in the characteristic indicator sample data to obtain M root cause classifications specifically includes:

[0026] An affinity propagation model is used to cluster all feature indicators in the feature indicator sample data to obtain the M root cause classifications.

[0027] Furthermore, the root cause classification includes: alarm class, coverage class, capacity class, interference class, structure class, access class, mobility class, retention class, and perception class.

[0028] Further, the comprehensive score of each root cause classification is calculated according to the number of all characteristic indicators and the number and importance of all key characteristic indicators in the M root cause classifications of the poor quality cells, specifically including:

[0029] According to the importance of all key characteristic indicators in the M root cause categories, obtaining the indicator importance ranking of each root cause category;

[0030] Obtain the priority ranking of each root cause classification based on expert experience;

[0031] For each root cause classification in the M root cause classifications, calculate the ratio of the number of all characteristic indicators in the root cause classification to the total number of characteristic indicators in all root cause classifications to obtain a first indicator number score for the root cause classification;

[0032] Calculate the second indicator score of each root cause classification according to the first indicator score of each root cause classification and the number of all key characteristic indicators in each root cause classification;

[0033] Scoring the second indicator number of each root cause classification to obtain an indicator number ranking of each root cause classification;

[0034] According to the indicator importance ranking, priority ranking and indicator number ranking of each root cause classification, the comprehensive score of each root cause classification is calculated.

[0035] Further, determining the main cause and the secondary cause of the problem of the poor quality cell according to the ranking of the comprehensive score specifically includes:

[0036] Determine the root cause classification ranked first in the comprehensive score as the preliminary main cause of the problem of the poor quality cell, and determine the root cause classification ranked second in the comprehensive score as the preliminary secondary cause of the problem of the poor quality cell;

[0037] Calibrate the characteristic indicators involved in the primary cause of the problem and the secondary cause of the problem of the poor quality cell through a preset value threshold range;

[0038] According to the characteristic indicators involved in the preliminary main cause and the preliminary secondary cause of the poor quality cell after the indicator calibration, updating the indicator importance ranking of each root cause classification;

[0039] Update the comprehensive score of each of the root cause classifications according to the updated indicator importance ranking of each of the root cause classifications, the priority ranking of each of the root cause classifications, and the indicator number ranking;

[0040] The root cause classification ranked first in the updated comprehensive score is determined as the final main cause of the problem of the poor quality cell, and the root cause classification ranked second in the updated comprehensive score is determined as the final secondary cause of the problem of the poor quality cell.

[0041] Further, after determining the main cause and the secondary cause of the problem of the poor quality cell according to the ranking of the comprehensive score, the method further includes:

[0042] According to the main causes and secondary causes of the problems of the poor quality cells and a preset network optimization case library, an optimization solution for the poor quality cells is obtained.

[0043] In a second aspect, the present invention provides a root cause analysis device for poor cell quality in a wireless network, comprising:

[0044] An acquisition module, used to acquire multi-dimensional characteristic indicator sample data of multiple wireless network cells;

[0045] An evaluation and screening module, connected to the acquisition module, for evaluating a plurality of classification models according to the characteristic index sample data to screen out the optimal classification model;

[0046] A selection module, connected to the evaluation and screening module, for selecting the top N key characteristic indicators with the highest impact on identifying poor-quality cells from the characteristic indicator sample data according to the optimal classification model, wherein 20≤N≤40;

[0047] A clustering module is connected to the selection module and is used to cluster all characteristic indicators in the characteristic indicator sample data to obtain M root cause classifications, where M <N;

[0048] A root cause analysis module is connected to the clustering module and is used to calculate a comprehensive score of each root cause classification for each poor-quality cell in the multiple wireless network cells according to the number of all characteristic indicators and the number and importance of all key characteristic indicators in the M root cause classifications of the poor-quality cell, and determine the main cause and secondary cause of the problem of the poor-quality cell according to the ranking of the comprehensive score.

[0049] In a third aspect, the present invention provides a root cause analysis device for a wireless network cell with poor quality, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to implement the root cause analysis method for a wireless network cell with poor quality as described in the first aspect above.

[0050] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for analyzing root causes of poor cell quality in a wireless network as described in the first aspect above is implemented.

[0051] The method, device and medium for analyzing the root cause of poor-quality wireless network cells provided by the present invention first obtain multi-dimensional characteristic index sample data of multiple wireless network cells; then evaluate multiple classification models according to the characteristic index sample data to screen out the optimal classification model; then according to the optimal classification model, select the top N key characteristic indexes with the highest influence on identifying poor-quality cells from the characteristic index sample data, where 20 ≤ N ≤ 40; then cluster all the characteristic indexes in the characteristic index sample data to obtain M root cause classifications, where M < N; finally, for each poor-quality cell among the multiple wireless network cells, calculate the comprehensive score of each root cause classification according to the number of all characteristic indexes, the number of all key characteristic indexes and the importance in the M root cause classifications of the poor-quality cell, and determine the main problem cause and secondary problem cause of the poor-quality cell according to the ranking of the comprehensive scores. By introducing the ideas of machine learning feature engineering, feature selection and importance determination, the present invention screens the optimal classification model based on multi-dimensional characteristic index sample data of multiple wireless network cells, extracts key characteristic indexes based on the optimal classification model, uses clustering problem rules to sort out M root cause classifications of poor-quality cells and calculate the comprehensive score of each root cause classification, and determines the main problem cause and secondary problem cause of poor-quality cells based on the ranking of the comprehensive scores of each root cause classification, so as to accurately locate wireless network problems, ultimately improve the quality of wireless network services and user perception, and solve the problem that the prior art generally relies on manual experience in analyzing the root cause of poor-quality wireless network cell problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a flowchart of a method for analyzing the root cause of poor-quality wireless network cells according to Embodiment 1 of the present invention;

[0053] Figure 2 It is a flowchart of another method for analyzing the root cause of poor-quality wireless network cells according to an embodiment of the present invention;

[0054] Figure 3 It is a sorting schematic diagram of an importance score example of characteristic indexes according to an embodiment of the present invention;

[0055] Figure 4 It is a schematic diagram of the clustering process of the affinity propagation model according to an embodiment of the present invention;

[0056] Figure 5 It is a flowchart of outputting the comprehensive score according to an embodiment of the present invention;

[0057] Figure 6 It is a structural schematic diagram of a device for analyzing the root cause of poor-quality wireless network cells according to Embodiment 2 of the present invention;

[0058] Figure 7This is a structural diagram of a device for analyzing root causes of poor cell quality in a wireless network according to embodiment 3 of the present invention. DETAILED DESCRIPTION

[0059] In order to enable those skilled in the art to better understand the technical solution of the present invention, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0060] It should be understood that the specific embodiments and drawings described herein are only used to explain the present invention rather than to limit the present invention.

[0061] It can be understood that, in the absence of conflict, the various embodiments of the present invention and the various features in the embodiments can be combined with each other.

[0062] It can be understood that, for the convenience of description, the drawings of the present invention only show the parts related to the present invention, while the parts irrelevant to the present invention are not shown in the drawings.

[0063] It can be understood that each unit and module involved in the embodiments of the present invention may correspond to only one physical structure, or may be composed of multiple physical structures, or multiple units and modules may be integrated into one physical structure.

[0064] It can be understood that, without conflict, the functions and steps marked in the flowcharts and block diagrams of the present invention may occur in an order different from that marked in the drawings.

[0065] It is understood that the flowcharts and block diagrams of the present invention illustrate the possible architectures, functions, and operations of the systems, devices, equipment, and methods according to the various embodiments of the present invention. Each box in the flowchart or block diagram may represent a unit, module, program segment, or code, which contains executable instructions for implementing the specified functions. Moreover, each box or combination of boxes in the block diagram and flowchart may be implemented by a hardware-based system that implements the specified functions, or may be implemented by a combination of hardware and computer instructions.

[0066] It can be understood that the units and modules involved in the embodiments of the present invention can be implemented by software or hardware. For example, the units and modules can be located in a processor.

[0067] Embodiment 1:

[0068] This embodiment provides a method for analyzing root causes of poor quality cells in a wireless network. Figure 1 As shown, the method includes:

[0069] Step S101: Acquire multi-dimensional characteristic indicator sample data of multiple wireless network cells.

[0070] It should be noted that the wireless network may refer to a 4G network or a 5G network, etc. The multi-dimensional characteristic indicator sample data includes network management performance data, measurement and collection data, fault alarm data, etc.

[0071] Step S102: Evaluate multiple classification models based on the characteristic index sample data to select the optimal classification model.

[0072] In this embodiment, based on multi-dimensional characteristic indicator sample data of multiple wireless network cells, different classification models are used for comparative evaluation, and the classification model with the best performance is selected as the model for subsequent use.

[0073] In an optional embodiment, the step of evaluating multiple classification models according to the characteristic index sample data to select the optimal classification model specifically includes:

[0074] Taking the multi-dimensional characteristic index sample data of each poor quality cell among the multiple wireless network cells as a positive sample, and taking the multi-dimensional characteristic index sample data of each non-poor quality cell among the multiple wireless network cells as a negative sample, to construct a classification training set;

[0075] Using the classification training set, respectively train the multiple classification models to obtain each of the classification models after training;

[0076] Calculate the final evaluation score of each classification model after training according to the accuracy of each classification model after training, the accuracy of balanced data, the area score under the receiver operating characteristic ROC curve, the F1 evaluation score and the training time;

[0077] The trained classification model with the highest final evaluation score is used as the optimal classification model.

[0078] Specifically, the multi-dimensional feature indicator sample data of each poor-quality cell is used as a positive sample, and the multi-dimensional feature indicator sample data of each non-poor-quality cell is used as a negative sample to construct a classification training set, and the classification training set is used to predict and evaluate different classification models. Each classification model is scored according to the accuracy of each trained classification model corresponding to the evaluation, the accuracy of the balanced data, the area score under the ROC curve (Receiver Operating Characteristic Curve), the F1 evaluation score and the training time, and the trained classification model with the highest final evaluation score is selected as the optimal classification model.

[0079] Specifically, the final evaluation score of each classification model after training is calculated by the following formula:

[0080] Final_Score=sum(Accuracy+Balance_Accuracy+Roc_Auc_Score

[0081] +F1_Score)-TimeTaken / 1000

[0082] Among them, Final_Score is the final evaluation score of the classification model, Accuracy is the accuracy of the classification model, Balance_Accuracy is the balanced data accuracy of the classification model, Roc_Auc_Score is the area score under the ROC curve of the classification model, F1_Score is the F1 evaluation score of the classification model, and TimeTaken is the training time of the classification model.

[0083] Step S103: According to the optimal classification model, the top N key characteristic indicators with the highest influence on identifying poor quality cells in the characteristic indicator sample data are selected, where 20≤N≤40.

[0084] Specifically, according to the training results of the optimal classification model, the importance of each feature indicator in the feature indicator sample data is determined, and the importance of the feature indicator is used to measure the influence of the feature indicator on the prediction target of the optimal classification model (identifying whether the wireless network cell is a poor quality cell); the importance of each feature indicator in the feature indicator sample data is arranged in descending order, and the first N feature indicators after descending order are selected as the first N key feature indicators with the highest influence on identifying poor quality cells as needed. As the underlying data basis for the output of the root cause analysis.

[0085] Step S104: clustering all characteristic indicators in the characteristic indicator sample data to obtain M root cause classifications, where M <N。

[0086] Specifically, an affinity propagation model is used to cluster all feature indicators in the feature indicator sample data according to feature indicator values ​​to obtain the M root cause classifications.

[0087] Step S105: For each poor quality cell among the multiple wireless network cells, a comprehensive score of each root cause classification is calculated according to the number of all characteristic indicators and the number and importance of all key characteristic indicators in the M root cause classifications of the poor quality cell, and the main cause and secondary cause of the problem of the poor quality cell are determined according to the ranking of the comprehensive score.

[0088] It should be noted that the root cause classification includes: alarm class, coverage class, capacity class, interference class, structure class, access class, mobility class, retention class, and perception class.

[0089] In an optional embodiment, the comprehensive score of each root cause classification is calculated according to the number of all characteristic indicators and the number and importance of all key characteristic indicators in the M root cause classifications of the poor quality cells, specifically including:

[0090] According to the importance of all key characteristic indicators in the M root cause categories, obtaining the indicator importance ranking of each root cause category;

[0091] Obtain the priority ranking of each root cause classification based on expert experience;

[0092] For each root cause classification in the M root cause classifications, calculate the ratio of the number of all characteristic indicators in the root cause classification to the total number of characteristic indicators in all root cause classifications to obtain a first indicator number score for the root cause classification;

[0093] Calculate the second indicator score of each root cause classification according to the first indicator score of each root cause classification and the number of all key characteristic indicators in each root cause classification;

[0094] Scoring the second indicator number of each root cause classification to obtain an indicator number ranking of each root cause classification;

[0095] According to the indicator importance ranking, priority ranking and indicator number ranking of each root cause classification, the comprehensive score of each root cause classification is calculated.

[0096] Specifically, the importance of all key feature indicators in each of the M root cause categories is accumulated, and based on the accumulated importance of all key feature indicators in each root cause category, the indicator importance ranking of each root cause category is obtained. The priority ranking of each root cause category based on expert optimization experience is obtained.

[0097] Specifically, the total number of characteristic indicators in all root cause classifications is calculated and divided by the number of all characteristic indicators in each of the M root cause classifications to obtain the first indicator number score of each root cause classification, and the first indicator number score of each root cause classification is calculated and multiplied by the number of all key characteristic indicators in each root cause classification to obtain the second indicator number score of each root cause classification and the indicator number ranking of each root cause classification.

[0098] Specifically, the comprehensive score of each root cause classification is obtained by accumulating the indicator importance ranking, priority ranking and indicator number ranking of each root cause classification.

[0099] In an optional embodiment, determining the main cause and the secondary cause of the problem of the poor quality cell according to the ranking of the comprehensive score specifically includes:

[0100] Determine the root cause classification ranked first in the comprehensive score as the preliminary main cause of the problem of the poor quality cell, and determine the root cause classification ranked second in the comprehensive score as the preliminary secondary cause of the problem of the poor quality cell;

[0101] Calibrate the characteristic indicators involved in the primary cause of the problem and the secondary cause of the problem of the poor quality cell through a preset value threshold range;

[0102] According to the characteristic indicators involved in the preliminary main cause and the preliminary secondary cause of the poor quality cell after the indicator calibration, updating the indicator importance ranking of each root cause classification;

[0103] Update the comprehensive score of each of the root cause classifications according to the updated indicator importance ranking of each of the root cause classifications, the priority ranking of each of the root cause classifications, and the indicator number ranking;

[0104] The root cause classification ranked first in the updated comprehensive score is determined as the final main cause of the problem of the poor quality cell, and the root cause classification ranked second in the updated comprehensive score is determined as the final secondary cause of the problem of the poor quality cell.

[0105] Specifically, the first and second root causes ranked in the comprehensive score are classified as the preliminary main causes and preliminary secondary causes of problems in poor quality cells. Statistical methods are combined with optimization experience to set the value threshold range for the characteristic indicators one by one, and the characteristic indicators related to the preliminary main causes and preliminary secondary causes of problems are calibrated according to the threshold configuration. The calibrated data is clustered again, and the indicator importance ranking is updated, and ranked jointly with the priority ranking and the indicator number ranking. The first one is the final main cause of the problem, and the second one is the final secondary cause of the problem.

[0106] In an optional embodiment, after determining the main cause and the secondary cause of the problem of the poor quality cell according to the ranking of the comprehensive score, the method further includes:

[0107] According to the main causes and secondary causes of the problems of the poor quality cells and a preset network optimization case library, an optimization solution for the poor quality cells is obtained.

[0108] Specifically, based on on-site optimization cases and the optimization experience of network experts, we sorted out the network optimization case library, and provided understandable optimization solutions by locating the main and secondary causes of problems in poor-quality cells and combining them with the network optimization case library.

[0109] In a specific embodiment, taking the wireless network as a 4G network as an example, Figure 2 As shown, the root cause analysis method of the poor quality cell in the wireless network includes the following steps:

[0110] Step S1: 4G network multi-dimensional data source fusion processing

[0111] Specifically, the multi-dimensional data of the 4G network wireless cell, including PM (network management performance data), MR (measurement and collection data), FM (fault alarm data), etc., is modeled according to business needs, and the data dictionary table supporting the upper-layer application is output. By extracting tables and indicators describing different business types, the multi-dimensional indicator performance table of the cell used for machine learning is automatically associated and output. The details are as follows:

[0112] The 4G cell performance hourly granularity data, 4G cell performance daily granularity data, 4G engineering parameters, and MR data are processed to obtain 4G cell characteristic indicators, and the indicator characteristics are classified and integrated in the cell dimension to form a multi-dimensional data source. The sample table of 4G wireless cell data structure is shown in Table 1:

[0113] Table 1: 4G wireless cell data structure sample table

[0114]

[0115]

[0116]

[0117]

[0118]

[0119] The above-mentioned poor quality data samples of the cells are fused to distinguish positive and negative samples. After completion, the process proceeds to step S2.

[0120] Step S2: Optimal model screening

[0121] Based on the various indicator values ​​and the quality difference label "problem_flag" fused from the multi-dimensional data source of the 4G cell in step S1, different models are used for comparative evaluation, and the model with the best performance is selected as the model for subsequent use. To select an appropriate classification model, accuracy, robustness, stability and performance must be considered comprehensively, so the optimal model is screened for a comprehensive score. In addition, due to the imbalance of the data, some unique evaluation indicators are adopted for evaluation. The evaluation method evaluates the effectiveness of the model from five dimensions, and the total score is the sum of the scores of each sub-item. Use problem_flag == 1 in the historical data of the same cell as the positive sample and problem_flag == 0 as the negative sample to construct a classification training set, use different models for prediction and evaluation, and score the model according to the evaluation. The evaluation items are as follows:

[0122] Accuracy: Accuracy measures the proportion of correctly classified samples among all prediction samples.

[0123]

[0124] Balanced_accuracy_score: A commonly used performance metric when dealing with imbalanced datasets. Unlike normal accuracy, balanced accuracy takes into account the imbalance in the number of samples between categories, so it is more suitable for model evaluation in the case of imbalanced data.

[0125]

[0126] Receiver Operating Characteristic Curve and Area Under ROC Curve Score (roc_auc_score): Receiver Operating Characteristic Curve (ROC Curve) and Area Under ROC Curve (AUC) are commonly used performance evaluation tools in binary classification problems to evaluate the accuracy and credibility of classification models. The ROC curve describes the relationship between the True Positive Rate (TPR, also known as the recall rate) and the False Positive Rate (FPR) at different classification thresholds. True Positive Rate TP is the number of true positive samples, P is the number of actual positive samples, and the false positive rate Where FP is the number of false positives, N is the number of actual negative samples, the horizontal axis of the ROC curve is FPR, and the vertical axis is TPR. AUC is the area under the ROC curve, ranging from 0 to 1. The larger the AUC, the better the classification performance of the model, and its ability to distinguish between positive and negative samples is stronger. When AUC is equal to 0.5, it means that the classification effect of the model is equivalent to random guessing; when AUC is greater than 0.5 and close to 1, it means that the classification effect of the model is better.

[0127] F1 Score: F1 evaluation is a performance indicator that comprehensively considers precision and recall.

[0128]

[0129] Training time (TimeTaken): Since the training time of different models varies greatly, and the root cause algorithm needs to be calculated for each cell, the time is an important evaluation indicator. Too long calculation time makes the model unusable, such as deep models.

[0130] Final score: The higher the score, the shorter the training time and the higher the model evaluation. Calculation rules:

[0131] Final_Score=sum(Accuracy+Balance_Accuracy+Roc_Auc_Score

[0132] +F1_Score)-TimeTaken / 1000

[0133] A large number of classification models were used in the model selection, and the final running results are shown in Table 2:

[0134] Table 2: Results of each classification model operation

[0135]

[0136] Through a series of evaluations based on live network data, the following model scores are output, and the optimal model is selected based on the scores: the model effectiveness is evaluated through five dimensions: accuracy, balanced data accuracy, area score under the ROC curve, F1 evaluation score, and training time. The higher the score, the shorter the training time, and the higher the model evaluation. Then, the classification model with the best comprehensive score is selected for feature selection training, and ExtraTreesClassifier is finally selected as the final model selection. After completion, proceed to step S3.

[0137] Step S3: Extraction of TOP N key features of 4G poor quality cells

[0138] Specifically, the historical data of the 4G cell is analyzed and modeled, and the model is established by using the quality difference label "problem_flag" == 0 or 1 as the target. Then the model will try its best to find the situation that can make the data item most likely to affect the result to analyze the feature weight. After the training of the optimal model screened out in step S2, the importance of the indicator features of each 4G cell is arranged in descending order, and the top 30 most influential features are selected as the underlying data basis for the root cause analysis output as needed. The TopN key features use the feature importance attribute of the AI ​​model. In the supervised model, feature importance refers to the degree of influence of each feature (input variable) in the model on the predicted target (output variable) in machine learning. In the extremely random tree, the evaluation is based on the Gini coefficient or information gain. In the construction process of each decision tree, the Gini coefficient or information gain can be calculated each time the node is split to measure the importance of each feature. The Gini coefficient and information gain are both used to measure the purity or uncertainty of the sample set. For each feature, its average Gini coefficient or average information gain can be calculated as a measure of feature importance. The Gini coefficient is an indicator used to measure the purity or uncertainty of a data set. For binary classification problems, the calculation of the Gini coefficient can be simplified as: Gini = 1-(p_1^2+p_2^2), where p1 is the proportion of samples belonging to category 1 to the total number of samples, and p2 = 1-p1 is the proportion of samples belonging to category 2 to the total number of samples. The value range of the Gini coefficient is between 0 and 1. The smaller the value, the higher the purity of the data set, that is, the more samples tend to belong to the same category. When the Gini coefficient is 0, it means that the samples in the data set belong to the same category and the data set is pure; when the Gini coefficient is 1, it means that the samples in the data set are evenly distributed in each category and the data set is impure. In the process of building a decision tree, selecting features with a smaller Gini coefficient for node division can make each node more pure, thereby building a more effective decision tree model. Through the extremely random tree model, after arranging each indicator feature in descending order according to the importance of the Gini factor, the top 30 most influential features are selected as the underlying data basis for the output of the root cause analysis as needed. An example of scoring the TOP important feature indicators (i.e. the importance of feature indicators) is as follows: Figure 3 As shown: After completion, go to step S4.

[0139] Step S4: Clustering of key features of 4G poor quality cells

[0140] Specifically, there are many indicators related to 4G wireless cells (there are thousands of indicators directly related), and it is extremely difficult for users to analyze so many indicators. If there is a method that can automatically classify approximate indicators and use a representative one (or several) indicators to represent this group of classifications, the threshold for use can be greatly reduced. In the present invention, this requirement can be met by the unsupervised clustering algorithm in the AI ​​algorithm, namely the Affinity Propagation Model, which is a clustering algorithm that determines the cluster center of the data point based on the similarity measure between the data points. By using the clustering algorithm, the following functions are completed: using the Affinity Propagation model to dynamically cluster key feature indicators; performing data verification on these classification indicators according to network optimization experience; manually adjusting the attribution of unreasonable feature indicators; and allocating weight coefficients according to the classification optimization processing priority.

[0141] 1. Algorithm advantages: 1) Automatic determination of cluster centers: The affinity propagation model is able to automatically select cluster centers without pre-specifying the number of clusters. It adapts to the characteristics of the data by propagating affinity between data points to determine which data points serve as cluster centers. 2) Applicable to irregular cluster shapes: Compared with traditional distance-based clustering algorithms (such as K-means), the affinity propagation model can adapt to irregular cluster shapes. It uses the concepts of affinity and attraction to establish complex relationships between data points, thereby capturing a richer clustering structure. 3) Not affected by the selection of initial values: The affinity propagation model is relatively insensitive to the choice of initial values. It uses message passing to update the affinity and attraction between data points, so it can adaptively adjust the clustering results and reduce dependence on initial values.

[0142] 2. Algorithm Disadvantages: 1) High time and space complexity: The affinity propagation model has a high computational overhead on large-scale data sets. It needs to calculate the affinity and attraction between each pair of data points, which results in a time complexity of O(N^2T), where N is the number of data points and T is the number of iterations. In addition, it also needs to store the affinity and attraction values ​​between each pair of data points, resulting in a high space complexity. 2) Sensitive to parameters: The affinity propagation model has some parameters that need to be set manually, such as the damping factor and the number of iterations. The choice of these parameters may have an impact on the clustering results, so parameter tuning is required to obtain better results. 3) Affected by noise and outliers: The affinity propagation model is sensitive to noise and outliers. Since it determines the cluster center based on the similarity measure between data points, if there is a lot of noise or outliers in the data set, it may lead to inaccurate clustering results.

[0143] The affinity model can cluster data into groups based on a large amount of data, so that the data in the same group represent the same meaning. The clustering process is as follows: Figure 4 Demonstration shown: Figure 4 In the figure, the affinity model clustering in the iterative process (ITERATION#1 to ITERATION#6) is shown, starting from initialization, after multiple iterations, until the final convergence (CONVERGENCE). At present, this process converges into a large category of problem classification, that is, according to the network optimization experience, the classification indicators are analyzed for business, the unreasonable characteristic indicator attribution is adjusted, and the characteristic indicator classification table is output. The affinity model clusters the 4G characteristic indicators according to the characteristic indicator values, assigns cluster numbers, and then analyzes the clustered clusters according to the network optimization experience. Finally, the 4G characteristic indicator classification configuration table is obtained as shown in Table 3:

[0144] Table 3: 4G characteristic index classification configuration table

[0145]

[0146]

[0147]

[0148]

[0149]

[0150] It can be seen from the data that there are indeed a lot of data that can be completely merged into one category. After the end, go to step S5.

[0151] Step S5: 4G poor quality cell root cause location

[0152] 1. If Figure 5 As shown in the figure, according to the three steps, the importance summing score of the top N key indicators of the cell (that is, the indicator importance ranking), the indicator classification weight coefficient score (that is, the indicator number ranking), and the indicator classification priority (that is, the priority ranking) score are completed. Finally, the 4G network cell feature classification priority comprehensive score (that is, the comprehensive score of the root cause classification) is output:

[0153] 1-1: The TOP indicator features (i.e. the top N key features) output by the indicator importance model are aggregated and ranked according to the root cause classification dimension. An example is shown in Table 4:

[0154] Table 4: TOP indicator characteristic root cause classification ranking table based on indicator importance

[0155]

[0156] The importance of these TOP indicators is aggregated according to the root cause classification dimension, and the importance ranking of each root cause classification of the cell is obtained according to the score ranking. As shown in the following table, the cell eci = 1461930906, there are 12 access items in the root cause classification, and the importance of these indicator items is accumulated, and the final calculation is 0.2789 points, ranking first, as shown in Table 5:

[0157] Table 5: Index importance ranking results based on index importance

[0158]

[0159] 1-2: Root cause classification and priority ranking are performed based on 4G cell characteristics obtained from expert optimization experience, as shown in Table 6:

[0160] Table 6: 4G cell root cause classification priority and corresponding score table based on expert experience

[0161]

[0162]

[0163] According to expert experience, the alarm category is given the highest priority for root cause analysis, so the score is full marks. The coverage category, capacity category, interference category, etc. are ranked second, third, and fourth respectively based on field experience, and so on. The ranking of all root cause categories and the corresponding score allocation are completed.

[0164] 1-3: According to the existing characteristic indicators and root cause classification, the number of indicators of each type of root cause is ranked, where the root cause classification indicator number score (that is, the first indicator number score) = the total number of characteristic indicators / the number of root cause classification indicators, as shown in Table 7:

[0165] Table 7: Root cause classification index number scoring

[0166] Root cause classification Root cause classification index number Total number of characteristic indicators Root cause classification index number scoring network Keep Class 18 194 10.78 4G Override Class 11 194 17.64 4G Perception 50 194 3.88 4G Interference 12 194 16.17 4G Alarm 6 194 32.33 4G Access 49 194 3.96 4G Structural Class 4 194 48.50 4G Capacity 15 194 12.93 4G Mobile 27 194 7.19 4G

[0167] It can be seen from the above table that the total number of characteristic indicators is the number of indicators in the "4G characteristic indicator classification configuration table" in step S4, among which the retention class has 18 indicators, the coverage class has 11 indicators, the perception class has 50 indicators, the interference class has 12 indicators, the alarm class has 6 indicators, the access class has 49 indicators, the structure class has 4 indicators, the capacity class has 15 indicators, and the mobility class has 27 indicators. According to the above calculation formula, the scoring results of the root cause classification indicators of different root cause classifications can be calculated.

[0168] Then, the following formula is used to complete the scoring of different indicator numbers. The formula is as follows: indicator number score (that is, the second indicator number score) = root cause classification indicator number score * the number of indicators of the root cause classification in the topN indicators. The indicator number score and ranking table are shown in Table 8:

[0169] Table 8: Indicator scoring and ranking table

[0170]

[0171]

[0172] According to the statistical results of the number of root cause classification indicators in the topN indicators of different root cause classifications of the cell eci=146193906 in Table 4, multiply it by the root cause classification indicator number scoring result in Table 7 to obtain the indicator number score.

[0173] Finally, according to the ranking of the three dimensions of indicator importance ranking, root cause classification priority ranking and indicator number ranking, the comprehensive ranking of 4G cell feature classification priority is obtained by accumulating the ranking of the three dimensions as shown in Table 9:

[0174] Table 9: 4G cell feature classification priority comprehensive ranking table

[0175] eci Root cause classification Index importance ranking Root cause classification priority ranking Ranking by number of indicators Overall Rating 146193906 Interference 3 3 1 7 146193906 Override Class 4 1 2 7 146193906 Structural Class 7 4 3 14 146193906 Access 2 5 4 11 146193906 Capacity 5 2 5 12 146193906 Perception 1 7 6 14 146193906 Mobile 6 6 7 19

[0176] 2. Defining the main and secondary causes of community problems

[0177] According to the results of the 4G cell feature classification priority comprehensive scoring system, the classification with a higher comprehensive ranking is selected as the basis for demarcating poor quality cells. The windowed ascending sorting label customization is performed according to the comprehensive ranking of the single cell problem classification. After associating the 4G cell problem list, the main cause (first problem) and secondary cause (second problem) of the 4G problem cell are output. As shown in Table 10, the interference class ranked first is the main cause, and the coverage class ranked second is the secondary cause.

[0178] Table 10: Sample results of the main and secondary causes of 4G problem cells

[0179]

[0180]

[0181] 3. Location of cell wireless root cause indicators

[0182] Analyze the characteristic indicators, use statistical methods combined with optimization experience, and set the value threshold range for the characteristic indicators one by one. Calibrate the indicators related to the main cause and secondary cause of the problem cell defined in the previous step (that is, the preliminary main cause of the problem and the preliminary secondary cause of the problem) according to the threshold configuration to obtain more accurate indicators related to the root cause of the problem. That is, output the main cause and secondary cause of the root cause positioning, and output the indicator values ​​related to the main cause and secondary cause. For the calibrated data, cluster the cell problems again, and rank them jointly with the problem classification priority and the number of indicators associated with the root cause of the problem. The first one is the main cause of the problem (that is, the final main cause of the problem), and the second one is the secondary cause of the problem (that is, the final secondary cause of the problem). For cells where only one root cause of the problem is identified, the main cause of the problem is the identified root cause of the problem, and the secondary cause of the problem is filled with "None". An example is shown in Table 11:

[0183] Table 11: 4G wireless cell root cause location example

[0184]

[0185] After completion, proceed to step S6.

[0186] Step S6: Output cell problem solution suggestions

[0187] Specifically, based on on-site optimization cases and the optimization experience of network experts, the network optimization case library is sorted out as the configuration of optimization suggestions, and the case content can be accumulated over a long period of time as the basis for the next knowledge base. As shown in Table 12:

[0188] Table 12: Network Optimization Case Library

[0189]

[0190]

[0191] Based on the root cause of the problem located, combined with the network optimization case library compiled by experts, a descriptive text is output to provide an understandable optimization solution. The example of the solution for 4G poor quality cells is shown in Table 13:

[0192] Table 13: 4G poor quality cell solution recommendations

[0193]

[0194]

[0195] After completion, proceed to step S7.

[0196] Step S7: End

[0197] It is worth mentioning that, in order to solve the above-described problems, the present invention proposes to incorporate the concept of AI intelligent analysis, establish an AI model for analyzing the root causes of poor-quality cell problems, realize the intelligent mining of the root causes of 4G network problems, and enhance the initiative in problem discovery. By introducing machine learning feature engineering, feature selection, and importance determination ideas, a correlation analysis is performed on multi-data sources such as network performance data, MR measurement data, interference data, and alarm data. Problem indicator features are extracted, and the TOP N reasons are sorted out using clustering problem rules to accurately locate 4G network problems, ultimately improving the service quality of the 4G network and user perception. By introducing concepts such as feature engineering and feature importance in AI, the present invention automatically analyzes the implicit root causes of a large amount of data. Problem symptoms are extracted, clustering problem rules are used, the TOP reasons for clustering are sorted out, the TOP factors are focused on, and complexity is simplified. Based on automation, data sources such as network performance data, MR measurement data, interference data, and alarm data are uniformly accessed. The main and secondary reasons are output for problem cells, network problems are accurately located, potential problem risk points are mined, platform experience is solidified, the construction of the problem library is improved, and optimization personnel are guided to handle network problems in a targeted manner. The work efficiency of front-line personnel is improved, the cost of network operation and maintenance is reduced, and the goal of "cost reduction and efficiency improvement" is effectively achieved.

[0198] The method for analyzing the root causes of poor-quality cells in a wireless network provided by an embodiment of the present invention first obtains multi-dimensional characteristic index sample data of multiple wireless network cells; then evaluates multiple classification models according to the characteristic index sample data to screen out the optimal classification model; then, according to the optimal classification model, selects the top N key characteristic indexes with the highest influence degree on identifying poor-quality cells from the characteristic index sample data, where 20 ≤ N ≤ 40; then clusters all the characteristic indexes in the characteristic index sample data to obtain M root cause classifications, where M < N; finally, for each poor-quality cell among the multiple wireless network cells, calculates the comprehensive score of each root cause classification according to the number of all characteristic indexes and the number and importance of all key characteristic indexes in the M root cause classifications of the poor-quality cell, and determines the main problem cause and secondary problem cause of the poor-quality cell according to the ranking of the comprehensive scores. By introducing machine learning feature engineering, feature selection, and importance determination ideas, the present invention screens out the optimal classification model based on multi-dimensional characteristic index sample data of multiple wireless network cells, extracts key characteristic indexes based on the optimal classification model, uses clustering problem rules to sort out M root cause classifications of poor-quality cells and calculates the comprehensive score of each root cause classification, and determines the main problem cause and secondary problem cause of the poor-quality cell based on the ranking of the comprehensive scores of each root cause classification, thereby accurately locating wireless network problems, ultimately improving the service quality of the wireless network and user perception, and solving the problem that the prior art generally relies on manual experience in analyzing the root causes of poor-quality cells in a wireless network.

[0199] Embodiment 2:

[0200] like Figure 6 As shown, this embodiment provides a root cause analysis device for a poor quality cell in a wireless network, which is used to execute the above-mentioned root cause analysis method for a poor quality cell in a wireless network, including:

[0201] An acquisition module 11 is used to acquire multi-dimensional characteristic indicator sample data of multiple wireless network cells;

[0202] An evaluation and screening module 12, connected to the acquisition module 11, is used to evaluate multiple classification models according to the characteristic index sample data to screen out the optimal classification model;

[0203] A selection module 13, connected to the evaluation and screening module 12, is used to select the top N key characteristic indicators with the highest influence on identifying poor quality cells from the characteristic indicator sample data according to the optimal classification model, wherein 20≤N≤40;

[0204] The clustering module 14 is connected to the selection module 13 and is used to cluster all the characteristic indicators in the characteristic indicator sample data to obtain M root cause classifications, where M <N;

[0205] The root cause analysis module 15 is connected to the clustering module 14, and is used to calculate the comprehensive score of each root cause classification for each poor quality cell in the multiple wireless network cells according to the number of all characteristic indicators and the number and importance of all key characteristic indicators in the M root cause classifications of the poor quality cell, and determine the main cause and secondary cause of the problem of the poor quality cell according to the ranking of the comprehensive score.

[0206] Furthermore, the evaluation and screening module 12 specifically includes:

[0207] A construction unit, configured to use the multi-dimensional characteristic indicator sample data of each poor-quality cell among the multiple wireless network cells as a positive sample, and use the multi-dimensional characteristic indicator sample data of each non-poor-quality cell among the multiple wireless network cells as a negative sample, to construct a classification training set;

[0208] A training unit, used to use the classification training set to train the multiple classification models respectively to obtain each of the classification models after training;

[0209] A first calculation unit is used to calculate a final evaluation score of each of the classification models after training according to the accuracy of each of the classification models after training, the accuracy of the balanced data, the area score under the receiver operating characteristic ROC curve, the F1 evaluation score and the training time;

[0210] As a unit, it is used to take the trained classification model with the highest final evaluation score as the optimal classification model.

[0211] Furthermore, the first computing unit is specifically configured to:

[0212] The final evaluation score of each classification model after training is calculated by the following formula:

[0213] Final_Score=sum(Accuracy+Balance_Accuracy+Roc_Auc_Score

[0214] +F1_Score)-TimeTaken / 1000

[0215] Among them, Final_Score is the final evaluation score of the classification model, Accuracy is the accuracy of the classification model, Balance_Accuracy is the balanced data accuracy of the classification model, Roc_Auc_Score is the area score under the ROC curve of the classification model, F1_Score is the F1 evaluation score of the classification model, and TimeTaken is the training time of the classification model.

[0216] Furthermore, the selection module 13 specifically includes:

[0217] A first determination unit, used to determine the importance of each feature indicator in the feature indicator sample data according to the training result of the optimal classification model, wherein the importance of the feature indicator is used to measure the influence of the feature indicator on the prediction target of the optimal classification model;

[0218] The arrangement and selection unit is used to arrange the importance of each characteristic indicator in the characteristic indicator sample data in descending order, and select the first N characteristic indicators after the descending order as the first N key characteristic indicators with the highest influence on identifying poor quality cells.

[0219] Furthermore, the clustering module 14 specifically includes:

[0220] A clustering unit is used to cluster all feature indicators in the feature indicator sample data using an affinity propagation model to obtain the M root cause classifications.

[0221] Furthermore, the root cause classification includes: alarm class, coverage class, capacity class, interference class, structure class, access class, mobility class, retention class, and perception class.

[0222] Furthermore, the root cause analysis module 15 includes:

[0223] A first obtaining unit is used to obtain an indicator importance ranking of each root cause classification according to the importance of all key feature indicators in the M root cause classifications;

[0224] An acquisition unit, used to acquire a priority ranking of each root cause classification obtained based on expert experience;

[0225] a calculating unit, configured to calculate, for each of the M root cause classifications, a ratio of the number of all characteristic indicators in the root cause classification to the total number of characteristic indicators in all the root cause classifications, to obtain a first indicator number score for the root cause classification;

[0226] A second calculation unit, configured to calculate a second indicator score for each of the root cause classifications according to the first indicator score for each of the root cause classifications and the number of all key characteristic indicators in each of the root cause classifications;

[0227] A second obtaining unit is used to score the second indicator number of each root cause classification to obtain the indicator number ranking of each root cause classification;

[0228] The third calculation unit is used to calculate the comprehensive score of each root cause classification according to the indicator importance ranking, priority ranking and indicator number ranking of each root cause classification.

[0229] Furthermore, the root cause analysis module 15 also includes:

[0230] A second determination unit is used to classify the root cause ranked first in the comprehensive score as the preliminary main cause of the problem of the poor quality cell, and classify the root cause ranked second in the comprehensive score as the preliminary secondary cause of the problem of the poor quality cell;

[0231] A calibration unit, configured to calibrate characteristic indicators involved in the primary cause of the problem and the secondary cause of the problem of the poor quality cell through a preset value threshold range;

[0232] A first updating unit is used to update the indicator importance ranking of each root cause classification according to the characteristic indicators involved in the preliminary main cause and the preliminary secondary cause of the poor quality cell after indicator calibration;

[0233] A second updating unit, configured to update the comprehensive score of each of the root cause classifications according to the updated indicator importance ranking of each of the root cause classifications, the priority ranking of each of the root cause classifications, and the indicator number ranking;

[0234] The third determination unit is used to determine the root cause classification ranked first in the updated comprehensive score as the final main cause of the problem of the poor quality cell, and determine the root cause classification ranked second in the updated comprehensive score as the final secondary cause of the problem of the poor quality cell.

[0235] Furthermore, the device also includes:

[0236] The obtaining module is used to obtain an optimization solution for the poor quality cell according to the main cause and secondary cause of the problem of the poor quality cell and a preset network optimization case library.

[0237] Embodiment 3:

[0238] refer to Figure 7 This embodiment provides a root cause analysis device for poor quality cells in a wireless network, including a memory 21 and a processor 22. The memory 21 stores a computer program, and the processor 22 is configured to run the computer program to execute the root cause analysis method for poor quality cells in a wireless network in Example 1.

[0239] The memory 21 is connected to the processor 22. The memory 21 may be a flash memory, a read-only memory or other memory. The processor 22 may be a central processing unit or a single-chip microcomputer.

[0240] Embodiment 4:

[0241] This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the root cause analysis method for poor quality cells in a wireless network in the above-mentioned embodiment 1 is implemented.

[0242] The computer-readable storage medium includes volatile or non-volatile, removable or non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, computer program modules or other data). Computer-readable storage media include, but are not limited to, RAM (Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable read only memory), flash memory or other memory technology, CD-ROM (Compact Disc Read-Only Memory), digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer.

[0243] In summary, the method, apparatus, and medium for analyzing the root cause of poor-quality wireless network cells provided by the embodiments of the present invention first obtain multi-dimensional characteristic index sample data of multiple wireless network cells; then evaluate multiple classification models according to the characteristic index sample data to select the optimal classification model; then, according to the optimal classification model, select the top N key characteristic indicators with the highest influence on identifying poor-quality cells from the characteristic index sample data, where 20 ≤ N ≤ 40; then cluster all the characteristic indicators in the characteristic index sample data to obtain M root cause classifications, where M < N; finally, for each poor-quality cell among the multiple wireless network cells, calculate the comprehensive score of each root cause classification according to the number and importance of all characteristic indicators and all key characteristic indicators in the M root cause classifications of the poor-quality cell, and determine the main problem cause and secondary problem cause of the poor-quality cell according to the ranking of the comprehensive scores. By introducing the ideas of machine learning feature engineering, feature selection, and importance determination, the present invention screens the optimal classification model based on the multi-dimensional characteristic index sample data of multiple wireless network cells, extracts key characteristic indicators based on the optimal classification model, uses clustering problem rules to sort out M root cause classifications of poor-quality cells and calculate the comprehensive score of each root cause classification, and determines the main problem cause and secondary problem cause of poor-quality cells based on the ranking of the comprehensive scores of each root cause classification, thereby accurately locating wireless network problems, ultimately improving the quality of wireless network services and user perception, and solving the problem that the prior art generally relies on manual experience in analyzing the root cause of poor-quality wireless network cell problems.

[0244] It can be understood that the above embodiments are merely exemplary embodiments adopted to illustrate the principle of the present invention, and the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also regarded as the protection scope of the present invention.

Claims

1. A method for analyzing root causes of poor quality cells in wireless networks, characterized in that: The method comprises: Obtain multi-dimensional characteristic indicator sample data of multiple wireless network cells; According to the characteristic indicator sample data, multiple classification models are evaluated to screen out the optimal classification model; According to the optimal classification model, the top N key characteristic indicators with the highest impact on identifying poor-quality cells in the characteristic indicator sample data are selected, where 20≤N≤40; Cluster all the characteristic indicators in the characteristic indicator sample data to obtain M root cause classifications, where M <N; For each poor quality cell among the multiple wireless network cells, a comprehensive score of each root cause classification is calculated according to the number of all characteristic indicators and the number and importance of all key characteristic indicators in the M root cause classifications of the poor quality cell, and the main cause and secondary cause of the problem of the poor quality cell are determined according to the ranking of the comprehensive score.

2. The method according to claim 1, characterized in that: The step of evaluating multiple classification models according to the characteristic index sample data to select the optimal classification model specifically includes: Taking the multi-dimensional characteristic index sample data of each poor quality cell among the multiple wireless network cells as a positive sample, and taking the multi-dimensional characteristic index sample data of each non-poor quality cell among the multiple wireless network cells as a negative sample, to construct a classification training set; Using the classification training set, respectively train the multiple classification models to obtain each of the classification models after training; Calculate the final evaluation score of each classification model after training according to the accuracy of each classification model after training, the accuracy of balanced data, the area score under the receiver operating characteristic ROC curve, the F1 evaluation score and the training time; The trained classification model with the highest final evaluation score is used as the optimal classification model.

3. The method according to claim 2, characterized in that The final evaluation score of each classification model after training is calculated according to the accuracy of each classification model after training, the accuracy of balanced data, the area score under the receiver operating characteristic ROC curve, the F1 evaluation score and the training time, specifically including: The final evaluation score of each classification model after training is calculated by the following formula: Final_Score=sum(Accuracy+Balance_Accuracy+Roc_Auc_Score+F1_Score)–TimeTaken / 1000 Among them, Final_Score is the final evaluation score of the classification model, Accuracy is the accuracy of the classification model, Balance_Accuracy is the balanced data accuracy of the classification model, Roc_Auc_Score is the area score under the ROC curve of the classification model, F1_Score is the F1 evaluation score of the classification model, and TimeTaken is the training time of the classification model.

4. The method according to claim 2, characterized in that: The selecting, according to the optimal classification model, the top N key characteristic indicators with the highest impact on identifying poor quality cells in the characteristic indicator sample data specifically includes: Determining the importance of each feature indicator in the feature indicator sample data according to the training result of the optimal classification model, wherein the importance of the feature indicator is used to measure the influence of the feature indicator on the prediction target of the optimal classification model; The importance of each characteristic indicator in the characteristic indicator sample data is arranged in descending order, and the first N characteristic indicators after the descending order are selected as the first N key characteristic indicators with the highest influence on identifying poor quality cells.

5. The method according to claim 1, characterized in that The clustering of all characteristic indicators in the characteristic indicator sample data to obtain M root cause classifications specifically includes: An affinity propagation model is used to cluster all feature indicators in the feature indicator sample data to obtain the M root cause classifications.

6. The method according to claim 1, characterized in that The root cause classification includes: alarm class, coverage class, capacity class, interference class, structure class, access class, mobility class, retention class, and perception class.

7. The method according to claim 1, characterized in that The comprehensive score of each root cause classification is calculated according to the number of all characteristic indicators and the number and importance of all key characteristic indicators in the M root cause classifications of the poor quality cells, specifically including: According to the importance of all key characteristic indicators in the M root cause categories, obtaining the indicator importance ranking of each root cause category; Obtain the priority ranking of each root cause classification based on expert experience; For each root cause classification in the M root cause classifications, calculate the ratio of the number of all characteristic indicators in the root cause classification to the total number of characteristic indicators in all root cause classifications to obtain a first indicator number score for the root cause classification; Calculate the second indicator score of each root cause classification according to the first indicator score of each root cause classification and the number of all key characteristic indicators in each root cause classification; Scoring the second indicator number of each root cause classification to obtain an indicator number ranking of each root cause classification; According to the indicator importance ranking, priority ranking and indicator number ranking of each root cause classification, the comprehensive score of each root cause classification is calculated.

8. The method according to claim 7, characterized in that Determining the main cause and the secondary cause of the problem of the poor quality cell according to the ranking of the comprehensive score specifically includes: Determine the root cause classification ranked first in the comprehensive score as the preliminary main cause of the problem of the poor quality cell, and determine the root cause classification ranked second in the comprehensive score as the preliminary secondary cause of the problem of the poor quality cell; Calibrate the characteristic indicators involved in the primary cause of the problem and the secondary cause of the problem of the poor quality cell through a preset value threshold range; According to the characteristic indicators involved in the preliminary main cause and the preliminary secondary cause of the poor quality cell after the indicator calibration, updating the indicator importance ranking of each root cause classification; Update the comprehensive score of each of the root cause classifications according to the updated indicator importance ranking of each of the root cause classifications, the priority ranking of each of the root cause classifications, and the indicator number ranking; The root cause classification ranked first in the updated comprehensive score is determined as the final main cause of the problem of the poor quality cell, and the root cause classification ranked second in the updated comprehensive score is determined as the final secondary cause of the problem of the poor quality cell.

9. The method according to claim 1, characterized in that: After determining the main cause and the secondary cause of the problem of the poor quality cell according to the ranking of the comprehensive score, the method further includes: According to the main causes and secondary causes of the problems of the poor quality cells and a preset network optimization case library, an optimization solution for the poor quality cells is obtained.

10. A root cause analysis device for poor quality cells in a wireless network, characterized in that: include: An acquisition module, used to acquire multi-dimensional characteristic indicator sample data of multiple wireless network cells; An evaluation and screening module, connected to the acquisition module, for evaluating a plurality of classification models according to the characteristic index sample data to screen out the optimal classification model; A selection module, connected to the evaluation and screening module, for selecting the top N key characteristic indicators with the highest impact on identifying poor-quality cells from the characteristic indicator sample data according to the optimal classification model, wherein 20≤N≤40; A clustering module is connected to the selection module and is used to cluster all the characteristic indicators in the characteristic indicator sample data to obtain M root cause classifications, where M <N; A root cause analysis module is connected to the clustering module and is used to calculate a comprehensive score of each root cause classification for each poor-quality cell in the multiple wireless network cells according to the number of all characteristic indicators and the number and importance of all key characteristic indicators in the M root cause classifications of the poor-quality cell, and determine the main cause and secondary cause of the problem of the poor-quality cell according to the ranking of the comprehensive score.

11. A root cause analysis device for poor quality cells in a wireless network, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to implement the root cause analysis method for poor cell quality in a wireless network as described in any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the root cause analysis method for poor cell quality in a wireless network according to any one of claims 1 to 9 is implemented.

Images