Root cause positioning method and device for key performance indicator anomalies

Abstract

The application provides a key performance indicator anomaly root cause positioning method and device, including using a machine learning model to select a labeled original data set, selecting a plurality of important performance indicators, and screening out associated performance indicators that do not participate in key performance indicator calculation. Based on the radio access network, the associated performance indicator data, the labeled key performance indicator data and the known root cause performance indicator data in multiple periods before and after the key performance indicator anomaly, a root cause positioning data set is constructed; the discrete associated performance indicator data in the root cause positioning data set is subjected to reason difference calculation to obtain the reason for the key performance indicator anomaly; the continuous associated performance indicator data in the root cause positioning data set is subjected to multiple correlation calculations to obtain the root cause of the key performance indicator anomaly. Through the above settings, the reason for the key performance indicator anomaly can be positioned from the discrete associated performance indicators, and the accuracy of root cause positioning is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the communication technology field, in particular to a method and device for locating the root cause of key performance indicator abnormality. BACKGROUND

[0002] A radio access network is a wireless implementation system for transmitting telecommunication services. Key performance indicators (KPIs) such as access type, maintenance type, mobility, and quality type represent the ability of the radio access network to provide services and can reflect the operation quality of the radio access network. When the radio access network coverage area has low radio access rate, high call drop rate, low handover success rate, and other key performance indicator abnormalities, the area can be determined as a poor quality area. Although key performance indicator abnormalities can identify poor quality areas, it is difficult to locate the root cause of poor quality because there are many reasons for poor quality of the radio access network.

[0003] In the prior art, the correlation between associated KPIs and main KPIs is calculated using Pearson or Spearman similarity calculation functions for root cause analysis, but Pearson or Spearman similarity calculation functions are not suitable for discrete data. Therefore, discrete correlation performance indicators are ignored in the root cause correlation analysis process, and the root cause analysis steps are not complete, which makes the root cause positioning inaccurate. SUMMARY

[0004] The present application provides a method and device for locating the root cause of key performance indicator abnormality, which aims to classify and calculate associated performance indicators and use multi-step calculation in correlation analysis to improve the accuracy of root cause positioning when key performance indicators are abnormal.

[0005] An embodiment of the present application provides a method for locating the root cause of key performance indicator abnormality, the method comprising:

[0006] Using a machine learning model to select the original data set that has been labeled, selecting a plurality of important performance indicators, and filtering the plurality of important performance indicators according to whether they participate in key performance indicator calculation to obtain associated performance indicators.

[0007] Based on the associated performance indicator data, the labeled key performance indicator data, and the known root cause performance indicator data in a plurality of periods before and after the key performance indicator of the radio access network becomes abnormal, a root cause positioning data set is constructed;

[0008] Reason difference calculation is performed on the discrete associated performance indicator data in the root cause positioning data set to obtain the cause of key performance indicator abnormality.

[0009] The correlation calculation is performed on the continuous correlation performance indicator data in the root cause positioning data set, and the root cause of the key performance indicator anomaly is obtained.

[0010] In an embodiment, the correlation calculation is performed on the continuous correlation performance indicator data in the root cause positioning data set, and the root cause of the key performance indicator anomaly is obtained, specifically including:

[0011] The correlation calculation is performed on the continuous correlation performance indicator data in the root cause positioning data set and the labeled key performance indicator data, and the continuous correlation performance indicator data with an absolute value of the correlation coefficient greater than or equal to a high correlation threshold value is obtained. The continuous correlation performance indicator data with the absolute value of the correlation coefficient greater than or equal to the high correlation threshold value is set as the abnormal reason indicator data.

[0012] The correlation calculation is performed on the known root cause performance indicator data and the abnormal reason indicator data in the root cause positioning data set, and the known root cause performance indicator data with an absolute value of the correlation coefficient greater than or equal to a high correlation threshold value is obtained. The known root cause performance indicator data with the absolute value of the correlation coefficient greater than or equal to the high correlation threshold value is the root cause of the key performance indicator anomaly.

[0013] In an embodiment, the reason difference calculation is performed on the discrete correlation performance indicator data in the root cause positioning data set, and the reason of the key performance indicator anomaly is obtained, specifically including:

[0014] For each discrete correlation performance indicator data, the difference between the mean of the discrete correlation performance indicator data in a plurality of periods after the key performance indicator anomaly and the mean of the discrete correlation performance indicator data in a plurality of periods before the key performance indicator anomaly is calculated.

[0015] The discrete correlation performance indicator with a difference greater than a preset difference threshold value is taken as the reason of the key performance indicator anomaly.

[0016] In an embodiment, before the correlation performance indicator is obtained, the method further includes:

[0017] Based on all the counters, performance indicators and key performance indicator data in the wireless access network communication process measured by the key performance indicator in the observation period, an original data set is constructed.

[0018] The key performance indicator data in the original data set is labeled according to the abnormal threshold value, the labeled key performance indicator data is obtained, and the labeled original data set is obtained.

[0019] In an embodiment, based on the correlation performance indicator data, the labeled key performance indicator data and the known root cause performance indicator data in a plurality of periods before and after the key performance indicator anomaly of the wireless access network, a root cause positioning data set is constructed, specifically including:

[0020] data conversion is performed on the associated performance indicator data, the labeled key performance indicator data and the known root cause performance indicator data to obtain an associated performance indicator feature vector, a labeled key performance indicator feature vector and a known root cause feature vector;

[0021] data preprocessing is performed on the associated performance indicator feature vector, the labeled key performance indicator feature vector and the known root cause feature vector;

[0022] attribute labeling is performed on the associated performance indicator feature vector according to a data type, and a root cause positioning dataset is constructed based on the associated performance indicator feature vector, the labeled key performance indicator feature vector and the known root cause feature vector; the data type includes discrete data and continuous data.

[0023] In an embodiment, the known root cause performance indicator data includes one or more combinations of uplink weak coverage, downlink weak coverage, over-coverage, uplink interference intensity and downlink interference rate.

[0024] Another embodiment of the present application provides a root cause positioning device for key performance indicator anomaly, comprising:

[0025] The feature acquisition module uses a machine learning model to select the labeled original dataset, selects a plurality of important performance indicators, and filters the plurality of important performance indicators according to whether they participate in key performance indicator calculation, and acquires associated performance indicators.

[0026] The index acquisition module is configured to construct a root cause positioning dataset based on the associated performance indicator data, the labeled key performance indicator data and the known root cause performance indicator data in a plurality of periods before and after the key performance indicator of the radio access network occurs anomaly.

[0027] The processing module performs cause difference calculation on the discrete associated performance indicator data in the root cause positioning dataset to obtain the cause of the key performance indicator anomaly.

[0028] The processing module performs multiple correlation calculations on the continuous associated performance indicator data in the root cause positioning dataset to obtain the root cause of the key performance indicator anomaly.

[0029] Another embodiment of the present application provides an electronic device, comprising a processor and a memory in communication with the processor;

[0030] The memory stores computer execution instructions;

[0031] The processor executes the computer execution instructions stored in the memory to implement the root cause positioning method for key performance indicator anomaly provided in the above embodiments.

[0032] A further embodiment of the present application provides a computer readable storage medium, wherein computer execution instructions are stored in the computer readable storage medium, and the computer execution instructions are used to implement the root cause positioning method of the key performance indicator exception provided in the above embodiment when executed by a processor.

[0033] A further embodiment of the present application provides a computer program product, comprising a computer program, which, when executed by a processor, implements the root cause positioning method of the key performance indicator exception provided in the above embodiment.

[0034] The present application provides a root cause positioning method and device of a key performance indicator exception, collects all counters, performance indicators and key performance indicator statistical data in a wireless access network communication process, generates an original data set, selects a number of most important features in the performance indicators according to a machine learning model, and excludes features participating in key performance indicator calculation according to prior knowledge, and the remaining features are associated performance indicators; the associated performance indicators, the key performance indicators and the known root cause performance indicator data in a plurality of periods before and after the key performance indicator exception are preprocessed, and a root cause positioning data set is constructed; meanwhile, the data types of the associated performance indicators in the root cause positioning data set are considered, and classification is calculated. The discrete associated performance indicators are calculated by difference, the failure reason of the key performance indicator exception is obtained, the continuous associated performance indicators are calculated by two-step correlation, and the root cause of the key performance indicator exception is output. When the key performance indicator is abnormal, not only can the reason be positioned from the discrete associated performance indicators, but also the root cause of the key performance indicator exception can be more accurately positioned according to the multi-step correlation calculation of the continuous key performance indicators. BRIEF DESCRIPTION OF DRAWINGS

[0035] The accompanying drawings, which are incorporated into and form part of the specification, illustrate embodiments consistent with the present application and, together with the specification, serve to explain the principles of the present application.

[0036] Figure 1 A flowchart of the root cause positioning method of the key performance indicator exception provided in an embodiment of the present application;

[0037] Figure 2 A flowchart of the root cause positioning method of the key performance indicator exception provided in another embodiment of the present application;

[0038] Figure 3 A flowchart of the root cause positioning method of the key performance indicator exception provided in a further embodiment of the present application;

[0039] Figure 4 A structural schematic diagram of the root cause positioning device of the key performance indicator exception provided in another embodiment of the present application;

[0040] Figure 5A structural schematic diagram of an electronic device provided for another embodiment of the present application.

[0041] The specific embodiments of the present application have been shown through the above-described drawings, and will be described in more detail hereinafter. These drawings and the written description are not intended to restrict the scope of the present application concept in any way, but to illustrate the present application concept to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0042] The exemplary embodiments will be described in detail herein with reference to the attached drawings. In the following description, the same numbers are used to indicate the same or similar components. The embodiments described in the following exemplary embodiments are not meant to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the present application as detailed in the appended claims.

[0043] The key performance indicator can reflect the quality of the service provided by the radio access network in the communication service process of the radio access network. When the key performance indicator is abnormal, it represents that the quality of the radio access network in the area is poor, and the root cause of the poor operation quality needs to be located.

[0044] When the KPI root cause correlation analysis method is used for positioning, the Pearson or Spearman similarity calculation function is only applicable to continuous data. In order to facilitate calculation, only discrete correlation performance indicators can be screened out when selecting correlation performance indicators. Moreover, the existing analysis steps are not perfect, so that the positioning accuracy of the KPI root cause correlation analysis method for the root cause is not high enough.

[0045] To solve the above technical problems, the present application provides a key performance indicator abnormality root cause positioning method and device, which aims to classify and calculate the correlation performance indicators and use multi-step calculation in correlation analysis to improve the accuracy of root cause positioning. The technical concept of the present application is that: collect all the counters, performance indicators and key performance indicator statistical data in the communication process of the radio access network, generate an original data set, use a custom model on the original data set, select a plurality of important features, and further screen the plurality of important features according to whether they participate in the calculation of the key performance indicator. Perform data preprocessing on the correlation performance indicators, key performance indicators and known root cause performance indicator data in multiple periods before and after the key performance indicator becomes abnormal, as a root cause positioning data set, and classify and calculate according to the data types in the root cause positioning data set. The discrete correlation performance indicators are subjected to difference value calculation to obtain the failure reason of the key performance indicator abnormality, and the continuous correlation performance indicators are subjected to two-step correlation calculation to output the root cause of the key performance indicator abnormality.

[0046] As Figure 1As shown, an embodiment of the present application provides a root cause positioning method for key performance indicator anomaly, which specifically comprises the following steps:

[0047] S101, use a machine learning model to select the labeled original data set, select a plurality of important performance indicators, and filter the plurality of important performance indicators according to whether they participate in the calculation of the key performance indicators, to obtain the associated performance indicators.

[0048] In this step, a machine learning model with feature importance (Feature Importance) measurement is selected, for example, a random forest with a decision tree-based learner. The performance indicators in the original data set are used as the input of the model, and the labeled key performance indicators in the original data set are used as the output of the model. The model is trained. The random forest can evaluate the importance of each performance indicator by comparing the contribution of each performance indicator in each tree in the random forest. A plurality of important performance indicators are selected. The performance indicators participating in the calculation of the key performance indicators are filtered out, for example, RRC connection establishment success rate * E-RAB establishment success rate = wireless connection rate. Therefore, the RRC connection establishment success rate and the E-RAB establishment success rate need to be filtered out, and the associated performance indicators are obtained.

[0049] S102, based on the key performance indicators of the wireless access network, the associated performance indicator data, the labeled key performance indicator data, and the known root cause performance indicator data in multiple periods before and after the anomaly, a root cause positioning data set is constructed.

[0050] In this step, the associated performance indicator data is obtained by selecting and filtering out the performance indicators participating in the calculation of the key performance indicators through a self-defined model. The key performance indicator data is labeled according to its abnormal threshold. When the key performance indicators of the wireless access network are abnormal, the associated performance indicator data, the labeled key performance indicator data, and the known root cause performance indicator data in multiple periods before and after the anomaly are obtained, which are used to construct the root cause positioning data set.

[0051] S103, reason difference calculation is performed on the discrete associated performance indicator data in the root cause positioning data set to obtain the reason for the key performance indicator anomaly.

[0052] Among them, the Pearson or Spearman similarity calculation function is not applicable to discrete associated performance indicator data, so the difference calculation method is adopted for discrete associated performance indicator data to obtain the reason for the key performance indicator anomaly.

[0053] S104, multiple correlation calculations are performed on the continuous associated performance indicator data in the root cause positioning data set to obtain the root cause of the key performance indicator anomaly.

[0054] In this step, the continuous correlation performance indicator data is first correlated with the key performance indicator data to obtain the reason performance indicator data with the highest correlation, and the traceability reasoning is completed. The reason performance indicator data with the highest correlation is correlated with the known root cause performance indicator data to calculate the root cause with the highest correlation, and the root cause of the key performance indicator abnormality is obtained.

[0055] In the above technical solution, the performance indicators in the communication process of the wireless access network record the operation of the wireless access network, and the correlation performance indicators can be screened out through the machine learning model and the judgment of whether to participate in the key performance indicator calculation. When the key performance indicator of the wireless access network is abnormal, based on the correlation performance indicator data, the labeled key performance indicator data and the known root cause performance indicator data in multiple periods before and after the abnormality, a root cause positioning data set is constructed. According to the data type of the correlation performance indicator in the root cause positioning data set, it is divided into continuous type and discrete type, the difference value of the discrete correlation performance indicator data is calculated, the key performance indicator abnormality reason is obtained, and the correlation of the continuous correlation performance indicator data is calculated multiple times to obtain the root cause of the key performance indicator abnormality.

[0056] As shown in Figure 2 An embodiment of the present application provides a root cause positioning method for key performance indicator abnormality. Before obtaining the correlation performance indicator, the method further includes the following steps:

[0057] S201, based on the key performance indicator data of all counters, performance indicators and key performance indicators in the communication process of the wireless access network measured in the observation period, an original data set is constructed.

[0058] In this step, the performance statistics data of the wireless access network is generated by a large number of counters, and different events and signaling messages in the communication process make the counters increase and decrease, and record the operation of the wireless access network. Various performance indicators are obtained by different counters. Among them, the performance indicators that can reflect the operation quality of the wireless access network are called key performance indicators, including access type, maintenance type, mobility, quality type, service integrity, utilization rate, availability and service type indicators. After the key performance indicator is abnormal, the original data set can be constructed according to the counters, performance indicators and key performance indicators in multiple periods before and after the abnormality.

[0059] S202, according to the abnormal threshold, the key performance indicators in the original data set are labeled to obtain the labeled key performance indicators, and the labeled original data set is obtained.

[0060] In this step, different key performance indicators correspond to different abnormal threshold values, and when the key performance indicator attribute value is greater than or less than the abnormal threshold value, the key performance indicator is marked, and the key performance indicator with an abnormal attribute value is marked as 1, and the key performance indicator with a normal attribute value is marked as 0. After marking the key performance indicators in the original data set, a marked original data set is obtained.

[0061] In the above technical solution, based on the original performance indicators of the key performance indicators in the wireless access network communication process in multiple weeks before and after the abnormality, an original data set is constructed. The key performance indicators in the original data set are marked and distinguished according to the abnormal threshold value, and a marked original data set is obtained.

[0062] An embodiment of the present application provides a root cause positioning method for key performance indicators. Based on the associated performance indicator data, the marked key performance indicator data and the known root cause performance indicator data in multiple cycles before and after the abnormality of the key performance indicators of the wireless access network, a root cause positioning data set is constructed, specifically including:

[0063] The known root cause performance indicator data includes one or more combinations of uplink weak coverage, downlink weak coverage, over-coverage, uplink interference intensity and downlink interference rate.

[0064] The known root cause performance indicator data contains all possible root causes of the key performance indicator abnormality, and the root cause of the key performance indicator abnormality needs to be accurately positioned.

[0065] The associated performance indicator data, the marked key performance indicator data and the known root cause performance indicator data are converted to obtain the associated performance indicator feature vector, the marked key performance indicator feature vector and the known root cause feature vector.

[0066] Each item of data is saved in the form of a feature vector to facilitate difference calculation and correlation calculation. For example, the associated performance indicator feature vector is x1 (i) , the known root cause feature vector is x2 (i) , and the marked key performance indicator feature vector is y (i) .

[0067] The associated performance indicator feature vector, the marked key performance indicator feature vector and the known root cause feature vector are preprocessed.

[0068] In the preprocessing, the feature vector with all empty values is deleted, the feature vector with missing values is filled with the mean value of the feature vector, the feature vector containing a percentage is converted to a real number, and the feature vector is standardized by standard deviation.

[0069] The correlation performance index feature vector is attribute-labeled according to a data type, and a root cause positioning data set is constructed based on the correlation performance index feature vector, the labeled key performance index feature vector, and a known root cause feature vector; the data type includes discrete data and continuous data.

[0070] The correlation performance index feature vector is divided into continuous and discrete types, the continuous correlation performance index feature vector includes uplink quality difference, downlink quality difference, uplink error code, downlink error code, RRC reconstruction, time delay, PRB utilization rate, signaling channel utilization rate, and 9 categories of switching, and each category includes multiple continuous feature vectors. The discrete correlation performance index feature vector is a signaling establishment failure abnormal reason, including the number of RRC connection establishment failures (UE non-response), the number of E_RAB establishment failures (core network problem, transmission layer problem), the number of RRC establishment congestion caused by wireless resource limitation; the number of E_RAB abnormal release (wireless layer problem, core network problem, base station problem, switching problem), the number of same frequency switching out failures, the number of eNodeB_switching failure UEContext release, the number of UE wireless connection loss leading to UE release, the number of connection user license limitation, and the length of LTE cell retreat service.

[0071] In the above technical solution, the correlation performance index data, the labeled key performance index data, and the known root cause performance index data are converted into the form of feature vectors, the feature vectors are preprocessed, the correlation performance index feature vectors are attribute-labeled according to discrete data and continuous data, and the root cause positioning data set is constructed according to the correlation performance index feature vector, the labeled key performance index feature vector, and the known root cause feature vector.

[0072] An embodiment of the present application provides a root cause positioning method for key performance index abnormality, which performs reason difference calculation on discrete index data in the root cause positioning data set to obtain the reason for key performance index abnormality, and specifically includes the following steps.

[0073] For each discrete correlation performance index data, the difference between the mean of the discrete correlation performance index data in a plurality of periods after the key performance index becomes abnormal and the mean of the discrete correlation performance index data in a plurality of periods before the key performance index becomes abnormal is calculated; and the discrete correlation performance index with a difference greater than a preset difference threshold is taken as the reason for key performance index abnormality.

[0074] For example, the number of times of E_RAB abnormal release (wireless layer problem) and the average number of RRC connections are discrete feature vectors. In the three periods before and after the abnormality of the key performance indicator, the average number of times of E_RAB abnormal release (wireless layer problem) in the last three periods is 170.67, the average number of times of E_RAB abnormal release (wireless layer problem) in the first three periods is 1.33, and the difference between the two is 169.34. The average number of RRC connections in the last three periods is 15.77, and the average number of RRC connections in the first three periods is 17.89, and the difference between the two is -2.12. The preset difference threshold is zero. Among the two discrete feature value vectors, only the difference between the last three periods and the first three periods of the number of times of E_RAB abnormal release is greater than zero, so the number of times of E_RAB abnormal release is output. Therefore, the number of times of E_RAB abnormal release (wireless layer problem) is the cause of the abnormality of the key performance indicator in the discrete correlation performance indicator.

[0075] In the above technical solution, the reason difference calculation is performed on the discrete correlation performance indicator data in the root cause positioning data set. The average of the discrete correlation performance indicator data in multiple periods after the abnormality of the key performance indicator occurs is used as the minuend, and the average of the discrete correlation performance indicator data in multiple periods before the abnormality of the key performance indicator occurs is used as the subtrahend. The difference between the averages of the discrete correlation performance indicator data in the multiple periods before and after is calculated. When the difference is greater than the preset difference threshold, the corresponding discrete correlation performance indicator is the cause of the abnormality of the key performance indicator.

[0076] As shown in Figure 3 An embodiment of the present application provides a root cause positioning method of a key performance indicator abnormality. Multiple correlation calculations are performed on the continuous correlation performance indicator data in the root cause positioning data set to obtain the root cause of the key performance indicator abnormality. Specifically, the method comprises the following steps:

[0077] In S301, correlation calculation is performed on the continuous correlation performance indicator data in the root cause positioning data set and the marked key performance indicator data to obtain continuous correlation performance indicator data with an absolute value of a correlation coefficient greater than or equal to a high correlation threshold. Continuous indicator data with an absolute value of a correlation coefficient greater than the high correlation threshold is set as abnormal cause indicator data.

[0078] In this step, the Pearson correlation coefficient between the continuous correlation performance indicator data and the labeled key performance indicator data is calculated, and the Pearson correlation coefficient value is between -1 and 1. When the Pearson correlation coefficient is positive, it indicates that the two data are positively correlated, and when the Pearson correlation coefficient is negative, it indicates that the two data are negatively correlated. According to the type of the continuous correlation performance indicator, the corresponding high correlation threshold is set, and when the absolute value of the Pearson correlation coefficient is greater than or equal to the high correlation threshold, and the continuous correlation performance indicator and the labeled key performance indicator data are positively or negatively correlated, the continuous correlation performance indicator is the cause of the key performance indicator anomaly. The feature vector with the largest absolute value of the Pearson correlation coefficient in each category of the continuous correlation performance indicator is set as the abnormal reason indicator data.

[0079] S302, the correlation between the known root cause performance indicator data in the root cause positioning data set and the abnormal reason indicator data is calculated, and the known root cause performance indicator data with an absolute value of the correlation coefficient greater than or equal to a high correlation threshold is obtained. The known root cause performance indicator data with an absolute value of the correlation coefficient greater than or equal to the high correlation threshold is the root cause of the key performance indicator anomaly.

[0080] In this step, the communication logical link is divided into uplink and downlink directions, and the Pearson correlation coefficient between the known root cause performance indicator data and the abnormal reason indicator data calculated in S301 in the same link direction is calculated. When the absolute value of the Pearson correlation coefficient is greater than or equal to the high correlation threshold, and the known root cause performance indicator data and the abnormal reason indicator data are positively or negatively correlated, the corresponding known root cause performance indicator data is the root cause of the key performance indicator anomaly.

[0081] For example: In the 3 periods before and after the key performance indicator call drop rate anomaly, the feature vector of the uplink initial HARQ retransmission ratio in the uplink error code category and the Pearson correlation coefficient of the key performance indicator are greater than the high correlation threshold and positively correlated, that is, the uplink initial HARQ retransmission ratio in the uplink error code category is the highest reason for the key performance indicator anomaly correlation. The known root cause performance indicator data and the uplink initial HARQ retransmission ratio feature vector in the uplink error code category are calculated for correlation, and it is found that the sampling point proportion of 0≤TA<0.5Km and the proportion of periodic reference signal received power greater than or equal to -105dBm are the root causes of the key performance indicator anomaly.

[0082] In the technical solution, the correlation of the continuous correlation performance indicator data in the root cause positioning data set is calculated in two steps. In the first step, the correlation of the continuous correlation performance indicator and the key performance indicator is calculated, the reason with the highest correlation with the key performance indicator is calculated, and the traceability reasoning is completed. In the second step, the correlation between the known root cause performance indicator and the correlation performance indicator with the highest correlation is calculated, the root cause with the highest correlation is calculated, and the root cause of the key performance indicator abnormality is obtained.

[0083] As shown in Figure 4 An embodiment of the present application provides a root cause positioning device 100 for key performance indicator abnormality, comprising:

[0084] The feature acquisition module 101 selects the labeled original data set using a machine learning model, selects a plurality of important performance indicators, and filters the plurality of important performance indicators according to whether they participate in key performance indicator calculation, and obtains correlation performance indicators.

[0085] The index acquisition module 102 is configured to construct a root cause positioning data set based on the correlation performance indicator data, the labeled key performance indicator data and the known root cause performance indicator data in a plurality of periods before and after the key performance indicator of the radio access network occurs abnormally.

[0086] The processing module 103 calculates the reason difference of the discrete correlation performance indicator data in the root cause positioning data set, and obtains the reason for the key performance indicator abnormality.

[0087] The processing module 103 performs multiple correlation calculations on the continuous correlation performance indicator data in the root cause positioning data set, and obtains the root cause of the key performance indicator abnormality.

[0088] In an embodiment, the feature acquisition module 101 is specifically configured to:

[0089] Based on the counter, performance indicator and key performance indicator data in the entire radio access network communication process measured by the key performance indicator in the observation period, an original data set is constructed;

[0090] The key performance indicator data in the original data set is labeled according to the abnormal threshold, and the labeled key performance indicator data is obtained.

[0091] The labeled original data set is selected using a machine learning model, a plurality of important performance indicators are selected, and the plurality of important performance indicators are filtered according to whether they participate in key performance indicator calculation, and correlation performance indicators are obtained.

[0092] In an embodiment, the index acquisition module 102 is specifically configured to:

[0093] The correlation performance indicator data, the labeled key performance indicator data and the known root cause performance indicator data are subjected to data conversion to obtain a correlation performance indicator feature vector, a labeled key performance indicator feature vector and a known root cause feature vector;

[0094] The correlation performance indicator feature vector, the labeled key performance indicator feature vector and the known root cause feature vector are subjected to data preprocessing;

[0095] The correlation performance indicator feature vector is subjected to attribute labeling according to a data type, and a root cause positioning dataset is constructed based on the correlation performance indicator feature vector, the labeled key performance indicator feature vector and the known root cause feature vector; the data type includes discrete data and continuous data.

[0096] In an embodiment, the processing module 103 is specifically configured to:

[0097] For each discrete correlation performance indicator data, a difference between a mean value of the discrete correlation performance indicator data in a plurality of periods after the abnormality occurs and a mean value of the discrete correlation performance indicator data in a plurality of periods before the abnormality occurs is calculated.

[0098] The discrete correlation performance indicator data with a difference greater than a preset difference threshold is set as a cause of the key performance indicator abnormality.

[0099] In an embodiment, the processing module 103 is specifically configured to:

[0100] The continuous correlation performance indicator data and the labeled key performance indicator data in the root cause positioning dataset are subjected to correlation calculation to obtain continuous correlation performance indicator data with an absolute value of a correlation coefficient greater than or equal to a high correlation threshold, and the continuous correlation performance indicator data with the absolute value of the correlation coefficient greater than or equal to the high correlation threshold is set as abnormal cause indicator data.

[0101] The known root cause performance indicator data and the abnormal cause indicator data in the root cause positioning dataset are subjected to correlation calculation to obtain known root cause performance indicator data with an absolute value of a correlation coefficient greater than or equal to a high correlation threshold, and the known root cause performance indicator data with the absolute value of the correlation coefficient greater than or equal to the high correlation threshold is a root cause of the key performance indicator abnormality.

[0102] As shown in FIG. 1, an embodiment of the present application provides an electronic device 200, which includes a memory 201 and a processor 202. Figure 5 The memory 201 is configured to store computer instructions executable by the processor.

[0103]

[0104] ​The processor 202 implements each step in the method in the above-mentioned embodiments when executing computer instructions. Details can be referred to the related description in the foregoing method embodiments.

[0105] Optionally, the memory 201 can be independent or integrated with the processor 202. When the memory 201 is independently arranged, the electronic device further includes a bus for connecting the memory 201 and the processor 202.

[0106] The embodiments of the present application further provide a computer readable storage medium, and the computer readable storage medium stores computer instructions. When the processor executes the computer instructions, each step in the method in the above-mentioned embodiments is implemented.

[0107] The embodiments of the present application further provide a computer program product, and the computer program product includes computer instructions. When the processor executes the computer instructions, each step in the method in the above-mentioned embodiments is implemented.

[0108] Other embodiments of the present application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. The application is intended to cover any variations, uses or adaptations of the application following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice in the art to which the application pertains. The specification and examples are to be regarded as illustrative only, and the true scope and spirit of the application are indicated by the following claims.

[0109] It should be understood that the application is not limited to the precise construction that has been described above and shown in the accompanying drawings, and that various modifications and changes can be made by those skilled in the art without departing from the scope of the application. The scope of the application is limited only by the appended claims.

Claims

1. A method for root cause localization of abnormal key performance indicators, characterized in that, include: A machine learning model is used to select multiple important performance indicators from the labeled original dataset. These important performance indicators are then filtered based on whether they participate in the calculation of key performance indicators (KPIs) to obtain related performance indicators. The related performance indicators do not include important performance indicators that participate in the calculation of KPIs. The labeled original dataset is obtained by labeling abnormal KPI data in the original dataset according to an anomaly threshold. Based on the correlation performance index data, the labeled key performance index data, and the known root cause performance index data within multiple periods before and after the anomaly of the key performance index of the wireless access network, a root cause localization dataset is constructed. For each discrete correlation performance index in the root cause localization dataset, calculate the difference between the mean of the discrete correlation performance index data in multiple periods after the key performance index becomes abnormal and the mean of the discrete correlation performance index data in multiple periods before the key performance index becomes abnormal; and take the discrete correlation performance index with a difference greater than a preset difference threshold as the cause of the key performance index abnormality. The correlation between the continuous correlation performance index data in the root cause localization dataset and the labeled key performance index data is calculated to obtain the continuous correlation performance index data with an absolute correlation coefficient greater than or equal to the high correlation threshold. The continuous correlation performance index data with an absolute correlation coefficient greater than or equal to the high correlation threshold is set as the abnormal cause index data. The known root cause performance index data and the abnormal cause index data in the root cause localization dataset are correlated to obtain known root cause performance index data whose absolute correlation coefficient is greater than or equal to the high correlation threshold. The known root cause performance index data whose absolute correlation coefficient is greater than or equal to the high correlation threshold is the root cause of the key performance index abnormality.

2. The root cause localization method according to claim 1, characterized in that, Before obtaining the relevant performance metrics, the method further includes: The original dataset is constructed based on all counters, performance indicators, and key performance indicator data in the wireless access network communication process measured by key performance indicators during the observation period.

3. The root cause localization method according to claim 1, characterized in that, Based on the correlation performance index data, labeled key performance index data, and known root cause performance index data across multiple periods before and after an anomaly in key performance indicators of the wireless access network, a root cause localization dataset is constructed, specifically including: Data transformation is performed on the associated performance index data, the labeled key performance index data, and the known root cause performance index data to obtain the associated performance index feature vector, the labeled key performance index feature vector, and the known root cause feature vector. Data preprocessing is performed on the feature vectors of the associated performance indicators, the feature vectors of the labeled key performance indicators, and the known root cause feature vectors. The associated performance index feature vectors are labeled with attributes according to the data type. Based on the associated performance index feature vectors, the labeled key performance index feature vectors, and the known root cause feature vectors, a root cause localization dataset is constructed. The data type includes discrete data and continuous data.

4. The root cause localization method according to claim 1, characterized in that, Known root cause performance metrics include one or more combinations of uplink weak coverage, downlink weak coverage, overcoverage, uplink interference intensity, and downlink interference rate.

5. A root cause localization device for abnormal key performance indicators, characterized in that, include: The feature acquisition module uses a machine learning model to select multiple important performance indicators from the labeled original dataset. It then filters these important performance indicators based on whether they participate in the calculation of key performance indicators to obtain related performance indicators. The related performance indicators do not include important performance indicators that participate in the calculation of key performance indicators. The labeled original dataset is obtained by labeling abnormal key performance indicator data in the original dataset according to an anomaly threshold. The indicator acquisition module is used to construct a root cause localization dataset based on the associated performance indicator data, the labeled key performance indicator data, and the known root cause performance indicator data within multiple periods before and after the occurrence of anomalies in the key performance indicators of the wireless access network. The processing module is used to calculate the difference between the mean of the discrete correlation performance index data in multiple periods after the key performance index becomes abnormal and the mean of the discrete correlation performance index data in multiple periods before the key performance index becomes abnormal for each discrete correlation performance index data in the root cause localization dataset; and to identify discrete correlation performance indexes whose difference is greater than a preset difference threshold as the cause of the key performance index abnormality. The processing module is used to perform correlation calculations on the continuous correlation performance index data in the root cause localization dataset and the labeled key performance index data, obtain continuous correlation performance index data with an absolute correlation coefficient greater than or equal to a high correlation threshold, and set the continuous correlation performance index data with an absolute correlation coefficient greater than or equal to the high correlation threshold as abnormal cause index data. The known root cause performance index data and the abnormal cause index data in the root cause localization dataset are correlated to obtain known root cause performance index data whose absolute correlation coefficient is greater than or equal to the high correlation threshold. The known root cause performance index data whose absolute correlation coefficient is greater than or equal to the high correlation threshold is the root cause of the key performance index abnormality.

6. An electronic device, comprising: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the root cause localization method for critical performance indicator anomalies as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the root cause localization method for critical performance indicator anomalies as described in any one of claims 1 to 4.

8. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 4.

Images