A KPI-based root cause positioning method, device, equipment and storage medium

Abstract

The application provides a KPI-based root cause positioning method, device, equipment and storage medium, the method comprising: for each target cell, calculating the correlation degree according to the sampling values of two different KPIs in the same time period, and determining the KPIs associated with each other; based on the sampling values of each KPI at different times and each influencing factor of the KPI, determining the contribution degree of each influencing factor to the KPI; based on different first importance evaluation indexes of each KPI, determining the importance of the KPI; based on the importance of any KPI, the correlation degree corresponding to the other KPIs associated with the KPI, and the contribution degree of each influencing factor of the KPI, determining the fault influence and the root cause of the fault. The application can efficiently and accurately locate the fault and reduce the fault recovery time, improve the operation and maintenance efficiency and user experience, etc.

Description

Technical Field

[0001] This application relates to the field of AI applications in communication technology, and in particular to a KPI-based root cause localization method, apparatus, device, and storage medium. Background Technology

[0002] Operations personnel monitor various Key Performance Indicators (KPIs) to pinpoint the KPIs causing problems and their specific causes, then resolve the issues and restore service. Therefore, accurate and efficient root cause analysis is crucial for improving operational efficiency and user experience.

[0003] Existing KPI root cause analysis methods can be summarized into three main categories: The first category requires operations personnel to manually search for problems, which is not only very time-consuming but also relies heavily on the experience of the operations personnel, making it difficult to achieve a uniform level of accuracy. The second category uses fault propagation graphs for root cause analysis, but currently, the information required to construct these graphs is insufficient, and the graphs change with service modifications, making maintenance difficult. The third category uses top-down or bottom-up search pruning anomaly detection methods to locate the root cause. However, in practical applications, if a misjudgment occurs at any stage, the computational workload for root cause analysis increases significantly, resulting in poor performance.

[0004] However, in actual 5G scenarios, due to the surge in the number of KPIs and their lack of periodicity and obvious regularity, existing KPI root cause analysis methods cannot effectively solve the problem of KPI root cause analysis in 5G scenarios. Therefore, a method is needed that can accurately and efficiently locate the problematic KPIs even in 5G scenarios, reduce fault recovery time, and improve operation and maintenance efficiency and user experience. Summary of the Invention

[0005] According to the KPI-based root cause localization method, apparatus, device and storage medium provided in this application, faults can be located quickly and efficiently, the specific KPIs that caused the faults and the root causes can be identified, and the fault recovery time can be reduced, and the operation and maintenance efficiency and user experience can be improved.

[0006] Firstly, this application provides a KPI-based root cause localization method, including:

[0007] For all KPIs of each target cell, the correlation between each pair of different KPIs is calculated based on the sampled values ​​of two different KPIs within the same time period, and the KPIs with correlation are determined.

[0008] For each target cell, based on the sampled values ​​of each KPI at multiple different times and the counters of each influencing factor affecting the KPI at the corresponding times, the contribution of each counter to the KPI is determined; based on the different first importance evaluation indicators of each KPI in each target cell, the importance of the KPI is determined.

[0009] For any target cell, the fault impact of a KPI is determined based on the importance of any KPI, the correlation of other KPIs related to that KPI, and the contribution of each counter that affects that KPI.

[0010] The root cause of the failure is determined by the failure impact of the KPIs of each target cell.

[0011] In one or more possible embodiments, the root cause of the fault is determined based on the fault impact of the KPIs of each target cell, including:

[0012] The top n KPIs with the largest fault impact are ranked according to the fault impact of each target cell.

[0013] The contribution of the counters for each of the first n KPIs is sorted, and the top m counters with the largest contribution to each KPI are identified as the root causes of the fault, where m and n are positive integers.

[0014] In one or more possible embodiments, it also includes:

[0015] The importance of each counter is determined based on the different second importance evaluation indicators of each counter in each target cell;

[0016] For any target cell, the fault impact of a counter is determined based on the importance of any counter and the contribution of that counter to each KPI.

[0017] In one or more possible embodiments, each KPI and counter is mapped to a node, establishing connections between nodes corresponding to two related KPIs, and connections between nodes corresponding to KPIs and counters with an impact relationship. The failure impact of the KPI / counter is determined, specifically including:

[0018] For any given node, the node's own influence is determined based on its importance and the first weighting factor. The influence of other nodes connected to the node on the node is determined based on the correlation / contribution of each of the other nodes connected to the node and the second weighting factor.

[0019] The failure impact of a node is determined based on the node's influence and the influence exerted on the node by other nodes connected to it.

[0020] Wherein, the first weighting factor is the reciprocal of the total number of nodes, and the second weighting factor is the failure impact of other nodes connected to the current node. In one or more possible embodiments, it further includes:

[0021] Map each KPI and counter to a node; connect the nodes corresponding to two related KPIs through a first edge, the weight of which is the degree of correlation between the two KPIs;

[0022] Connect the nodes corresponding to the KPIs and counters that have an influence relationship through a second edge to obtain the KPI association propagation graph. The weight of the second edge is the contribution of the counter to the KPI.

[0023] In one or more possible embodiments, it also includes:

[0024] Obtain at least one key performance indicator (KPI) for each of the multiple cells;

[0025] When it is determined that any of the acquired key KPIs has a preset threshold, the key KPI is compared with the corresponding preset threshold to determine whether the key KPI is abnormal.

[0026] If any key KPI is determined to be outside the preset threshold, a probabilistic statistical model is used to predict whether the key KPI is abnormal; target cells with abnormal key KPIs are then selected.

[0027] In one or more possible embodiments, a probabilistic statistical model is used to predict whether the key KPI is abnormal, including:

[0028] Input all sampled values ​​of any key KPI that does not have a preset threshold into a probabilistic statistical model; use the probabilistic statistical model to aggregate all sampled values ​​into two categories, and determine the anomaly category based on the number of sampled values ​​in each category;

[0029] Determine whether the key KPI is abnormal based on the category to which the current sampled value of the key KPI belongs.

[0030] In one or more possible embodiments, the method further includes: evaluating the correctness of the probabilistic statistical model, and determining the accuracy of the results of whether the key KPIs are abnormal based on the correctness evaluation results of the probabilistic model.

[0031] In one or more possible embodiments, the contribution of each counter to the KPI is determined using a linear fitting method, including:

[0032] By taking each KPI at multiple times as a vector y and taking each counter that affects the KPI at the corresponding time as a vector x, multiple discrete points are obtained.

[0033] The partial least squares method is used to perform linear fitting on the multiple discrete points to obtain the vector k in the linear regression curve y = kx. The vector k is then determined as the contribution of the counter to the KPI.

[0034] In one or more possible embodiments, the first importance evaluation metric includes any one or more of the following:

[0035] Whether it is a key KPI, whether it is an abnormal KPI, and the first initial weight set for different non-key KPIs.

[0036] In one or more possible embodiments, the second important evaluation index is a second initial weight set for different counters, or the weight of each counter is calculated based on whether the KPIs related to each counter are abnormal.

[0037] In one or more possible embodiments, the correlation between each pair of different KPIs is calculated using a normalized cross-correlation method to determine the correlated KPIs, specifically including:

[0038] The correlation between any two different KPIs is calculated using the normalized cross-correlation method, and two KPIs with a correlation greater than a set threshold are identified as related KPIs.

[0039] In one or more possible embodiments, the key KPIs include any one or more of the following:

[0040] Wireless access success rate, UE context establishment success rate, RRC connection establishment success rate, and RRC reconstruction success rate.

[0041] Secondly, this application also provides a KPI-based root cause localization device, comprising:

[0042] The correlation determination module is used to calculate the correlation between any two different KPIs based on the sampled values ​​of two different KPIs at the same time for all KPIs in each target cell, and to determine the KPIs that are correlated.

[0043] The contribution determination module is used to determine the contribution of each counter to the KPI for each target cell based on the sampled values ​​of each KPI at different times and the counters of each influencing factor affecting the KPI at the corresponding times.

[0044] The importance determination module is used to determine the importance of each KPI based on the different first importance evaluation indicators for each KPI in each target cell;

[0045] The impact determination module is used to determine the fault impact of any target cell based on the importance of any KPI, the correlation of other KPIs related to this KPI, and the contribution of each counter that affects this KPI.

[0046] The fault determination module is used to determine the root cause of a fault based on the fault impact of the KPIs of each target cell.

[0047] Thirdly, this application also provides a KPI-based root cause localization device, the device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the KPI-based root cause localization method as described in any of the first aspects.

[0048] Fourthly, this application also provides a storage medium storing a computer program for causing a computer to perform the KPI-based root cause localization method as described in any of the first aspects.

[0049] According to the KPI-based root cause localization method, apparatus, device and storage medium provided in this application, faults can be located quickly and efficiently, the specific KPIs that caused the faults and the root causes can be identified, and the fault recovery time can be reduced, and the operation and maintenance efficiency and user experience can be improved. Attached Figure Description

[0050] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application, and do not constitute an undue limitation of this application.

[0051] Figure 1 A flowchart provided according to one embodiment of this application;

[0052] Figure 2 A flowchart provided according to one embodiment of this application;

[0053] Figure 3 A flowchart provided according to one embodiment of this application;

[0054] Figure 4 This is a schematic diagram illustrating an anomaly of a key KPI provided according to one embodiment of this application;

[0055] Figure 5This is a KPI correlation propagation graph provided according to one embodiment of this application;

[0056] Figure 6 This is a schematic diagram illustrating the correlation between two KPIs according to an embodiment of this application;

[0057] Figure 7 This is a schematic diagram of an apparatus provided according to an embodiment of this application;

[0058] Figure 8 This is a schematic diagram of a device according to an embodiment of this application;

[0059] Figure 9 This is a schematic diagram of a computer storage medium provided according to an embodiment of this application. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0061] Existing technical solutions for fault root cause localization generally fall into two main categories. The first category uses top-down or bottom-up search-pruning anomaly detection methods to locate the root cause. In practical applications, if a misjudgment occurs at any stage, the computational load for root cause localization increases significantly, resulting in poor performance. Furthermore, it's difficult to guarantee the accuracy of anomaly detection, as it doesn't consider the correlation between KPIs. In 5G scenarios, many KPIs are not periodic and lack clear patterns, making it difficult to predict anomalies in specific points or intervals based on historical trends. Some methods in this category use prediction rules, which are unlikely to cover most faults. In terms of efficiency, assuming a dataset has 10 KPIs, each with 5 dimensions, there are 5 to the power of 10 (9,765,625) different combinations. Manually analyzing these combinations is time-consuming and easily influenced by experience. In 5G scenarios, there are tens of thousands of KPIs, each with dozens or hundreds of dimensions. Even with tree-based pruning search, the efficiency remains low, and it's difficult to guarantee the accuracy of anomaly detection. The second category involves using knowledge graphs or machine learning to establish causal relationships between KPIs. However, in reality, there are not many causal relationships between KPIs. Actual business changes rapidly, making it difficult to capture causal relationships. Although this approach overcomes the inefficiency of tree-based retrieval and considers the correlation between KPIs, the results are not very good due to the limited richness of the knowledge graph.

[0062] To address the aforementioned problems, this application provides a KPI-based root cause localization method, such as... Figure 1 As shown, it includes:

[0063] Step 101: For all KPIs of each target cell, calculate the correlation between each pair of different KPIs based on the sampled values ​​of two different KPIs within the same time period, and determine the KPIs that are related.

[0064] All the above KPIs include critical KPIs and non-critical KPIs. For the sampled values ​​of any two different KPIs in each target cell within the same time period, the correlation between any two different KPIs can be calculated using the normalized cross-correlation method. Since there are KPI sampled values ​​at multiple times within a certain period, the KPI sampled values ​​at multiple sampling times can be combined, and the correlation between any two different KPIs can be calculated using the normalized cross-correlation method.

[0065] Step 102: For each target cell, based on the sampled values ​​of each KPI at different times and the corresponding counters of the factors affecting the KPI at those times, determine the contribution of each counter to the KPI.

[0066] For different KPIs, there may be one or more counter factors affecting that KPI. One counter may affect one or more KPIs. KPIs and counters can be predefined. During root cause analysis, for each KPI within a certain period, the corresponding KPI and counter at each sampling time are determined, thus obtaining a discrete point. Since there is a linear relationship between KPIs and counters, the linear fitting method can be used to determine the linear relationship between KPIs and counters, thereby determining the contribution of each counter to the KPI.

[0067] Step 103: Determine the importance of each KPI based on the different first importance evaluation indicators for each KPI in each target cell.

[0068] For KPIs, at least one first importance evaluation indicator can be predefined. Since KPIs can be divided into key KPIs and non-key KPIs, corresponding first importance evaluation indicators can be defined for key KPIs and non-key KPIs respectively. The above first importance evaluation indicators include any one or more of the following: whether it is a key KPI, whether it is an abnormal KPI, and the first initial weight set for different non-key KPIs respectively.

[0069] Step 104: For any target cell, determine the fault impact of the KPI based on the importance of any KPI, the correlation degree of other KPIs related to the KPI, and the contribution of each counter affecting the KPI.

[0070] The importance of the aforementioned KPI, the correlation with other KPIs related to it, and the contribution of each counter affecting the KPI are all related to the failure impact of the KPI. For example, the higher the importance of the KPI, the greater its failure impact; the higher the correlation with other KPIs related to it, the greater its failure impact; and the greater the contribution of each counter affecting the KPI, the greater its failure impact. Therefore, the failure impact of the KPI can be determined by comprehensively considering the above factors.

[0071] Step 105: Determine the root cause of the fault based on the fault impact of the KPIs of each target cell.

[0072] This embodiment comprehensively considers the impact of the above factors, including the influence of the importance of the KPI itself on the KPI's failure impact, the influence of the correlation between the KPI and other KPIs associated with the KPI on the KPI's failure impact, and the influence of the contribution of the KPI to the KPI's failure impact due to the KPI having one or more counters. This allows for direct identification of KPIs with significant failure impact, enabling rapid and efficient accurate fault location, identification of the specific KPI causing the failure and its root cause, reduction of fault recovery time, and improvement of operational efficiency and user experience.

[0073] This application embodiment also includes determining the specific implementation method for each target cell, such as... Figure 2 As shown, it includes the following steps:

[0074] Step 201: Obtain at least one key performance indicator (KPI) corresponding to each of the multiple cells;

[0075] In one or more possible embodiments, when obtaining at least one key performance indicator (KPI) corresponding to multiple cells, key KPIs and non-key KPIs are determined by pre-classifying and classifying the KPIs. Key KPIs are also the causes that significantly impact faults, which can effectively improve the efficiency of filtering out abnormal KPIs. The aforementioned key KPIs include any one or more of the following: radio access success rate, user equipment (UE) context establishment success rate, radio resource control (RRC) connection establishment success rate, and RRC reconstruction success rate. These key KPIs are generally in the range of dozens to hundreds, representing the operator's or company's classification and definition of KPIs, and include, but are not limited to, any one or more of the above.

[0076] The radio access success rate reflects the UE access quality. Related statistics are performed at the Centralized Unit (CU) RRC layer. Within the statistical period, the radio access success rate is calculated by multiplying the RRC connection establishment success rate by the NG interface signaling connection success rate between the radio access network and the 5G core network, and the initial Quality of Service (QoS) flow establishment success rate. The UE context establishment success rate reflects the UE access quality. Related statistics are performed at the CU RRC layer. Within the statistical period, the UE context establishment success rate is calculated as the ratio of the number of successful UE context establishments to the number of UE context establishment requests. The RRC connection establishment success rate reflects the UE access quality. Related statistics are performed at the CU RRC layer. Within the statistical period, the RRC connection establishment success rate is calculated as the ratio of the number of successful RRC connection establishments to the number of RRC connection establishment requests. The RRC reconstruction success rate reflects the UE service maintenance quality. Related statistics are performed at the CU RRC layer. Within the statistical period, the RRC reconstruction success rate is calculated as the ratio of the number of completed RRC reconstructions to the number of RRC reconstruction requests.

[0077] Step 202: When it is determined that any of the acquired key KPIs has a preset threshold, the key KPI is compared with the corresponding preset threshold to determine whether the key KPI is abnormal.

[0078] For certain key KPIs, if the value of the key KPI is greater than a preset threshold, the key KPI is determined to be abnormal. For certain key KPIs, if the value of the key KPI is less than a preset threshold, the key KPI is determined to be abnormal.

[0079] like Figure 4 As shown, the threshold for a certain key KPI is 99. If the value of the key KPI is less than 99.0, it is determined to be an abnormal key KPI.

[0080] In one or more possible embodiments, when using the aforementioned preset threshold to determine whether a key KPI is abnormal, one can choose to comprehensively calculate all sampled values ​​within a certain period of time according to the calculation formula of the corresponding KPI defined by the operator, and use the sampled values ​​of the key KPI within a certain period of time as a whole to determine whether the key KPI is abnormal; or, one can directly compare the sampled value of a key KPI at a certain moment with the corresponding preset threshold to determine whether the key KPI is abnormal at the current moment.

[0081] Step 203: If any of the acquired key KPIs does not have a preset threshold, use a probability statistical model to predict whether the key KPI is abnormal.

[0082] Probabilistic statistical models can classify key KPIs that do not currently have preset thresholds based on large amounts of data. For different key KPIs, a corresponding probabilistic statistical model can be used to predict whether the key KPI is abnormal.

[0083] Step 204: Filter out target cells with abnormal key KPIs.

[0084] For a given target cell, if there is an abnormal key performance indicator (KPI), then the target cell is identified as the target cell with the abnormal key performance indicator.

[0085] In one or more possible implementations, the root cause of a fault is determined based on the fault impact of the KPIs of each target cell. This includes: ranking the KPIs of each target cell according to their fault impact to determine the top n KPIs with the largest fault impact; the top n KPIs are the causes of the fault, and the contribution of the counters of each of the top n KPIs is ranked to determine the top m counters with the largest contribution to each KPI as the root cause of the fault. Since the causes of KPI faults are also different, the root cause of the fault can be basically determined based on the contribution of each counter to the KPI; where m and n are positive integers.

[0086] In one or more possible embodiments, the method further includes: determining the importance of each counter based on different second importance evaluation indicators for each counter in each target cell; and, for any target cell, determining the fault impact of a counter based on the importance of any counter and the contribution of that counter to each KPI it affects. The aforementioned second importance evaluation indicators are either second initial weights set separately for different counters, or the weights of each counter are calculated based on whether each counter's impact on the affected KPIs is abnormal.

[0087] The aforementioned second initial weight is a fixed importance level. Regardless of its relationship with abnormal KPIs, the importance of all counters is the set initial value. Alternatively, the importance of counters can be calculated as non-critical KPIs, which could be 0.6*0.3+0.4*0.8=0.5 or 0.6*0.3+0.4*0.2=0.26. Based on the importance of any counter and its contribution to the various KPIs it affects, the failure impact of that counter is determined.

[0088] In one or more possible embodiments, each KPI and counter is mapped to a node, establishing connections between nodes corresponding to two related KPIs, and connections between nodes corresponding to KPIs and counters with an influence relationship. The failure impact of the KPI / counter is determined, specifically including: for any node, determining its own influence based on its importance and a first weighting factor; determining the influence of other nodes connected to the node based on the correlation / contribution of all other nodes connected to the node and a second weighting factor; and determining the failure impact of the node based on its influence and the influence of other nodes connected to the node. The first weighting factor is the reciprocal of the total number of nodes, and the second weighting factor is the failure impact of other nodes connected to the node. The specific calculation formula is as follows:

[0089]

[0090] NR i Let N represent the impact of a failure on the i-th node, and N represent the number of nodes. IMT i Let Ln(V) represent the importance of the i-th node. C is a constant used to adjust the proportion of influence given by the node itself and other nodes connected to it. i ) represents the set of all pointers to node i, Out(V j ) represents the set of points to other nodes for node j. W ji This represents the correlation or contribution between node j and node i. In the formula above, the left half represents the node's own influence, and the right half represents the influence given by other nodes connected to that node. Therefore, the fault influence of each node can be determined. By ranking the fault influence of each node, several KPIs with a significant impact on the fault can be identified. Furthermore, by ranking the counters corresponding to these KPIs with significant impact, the specific cause of the fault can be determined. For example, ranking the KPIs by their fault influence determines that the RRC connection establishment success rate is the KPI causing the fault. Ranking the counters corresponding to the RRC connection establishment success rate by their contribution determines that one or more of the following counters—RRC connection establishment success rate, the ratio of RRC connection establishment requests, admission failures, air interface timer timeouts, cell rejections, and other reasons—are the root cause of the fault.

[0091] In one or more possible embodiments, such as Figure 5As shown, it also includes: mapping each KPI and counter to nodes; connecting the nodes corresponding to two related KPIs through a first edge, the weight of which is the correlation between the two KPIs; connecting the nodes corresponding to KPIs and counters with an influence relationship through a second edge to obtain a KPI correlation propagation graph, the weight of which is the contribution of the counter to the KPI. This makes it easier to view the connection relationships between each KPI and other KPIs and counters, making the connection relationships between nodes more concise and clear, and facilitating quick problem identification. In one or more possible embodiments, a probabilistic statistical model is used to predict whether the key KPI is abnormal, such as... Figure 3 As shown, it includes:

[0092] Step 301: Input all sampled values ​​of any key KPI that does not have a preset threshold into the probability statistical model;

[0093] Step 302: Using the probabilistic statistical model, all sampled values ​​are aggregated into two categories, and the anomaly category is determined based on the number of sampled values ​​in each category;

[0094] Step 303: Determine whether the key KPI is abnormal based on the category to which the current sampled value of the key KPI belongs.

[0095] In one or more possible embodiments, a probabilistic statistical model is used to predict whether a key KPI is abnormal. This includes: inputting all sampled values ​​of any key KPI that does not have a preset threshold into the probabilistic statistical model; using the probabilistic statistical model to aggregate all sampled values ​​into two categories, and determining the abnormal category based on the number of sampled values ​​in each category; and determining whether the key KPI is abnormal based on the category in which the current sampled value of the key KPI belongs. When using a probabilistic statistical model to determine whether a key KPI is abnormal, it is also possible to choose to judge whether the sampled values ​​of the key KPI within a certain period of time or the sampled values ​​at a certain moment are abnormal, depending on specific business requirements.

[0096] In one or more possible embodiments, the method further includes: classifying key KPI anomalies based on a general statistical model used to predict key KPIs within a historical time period, and classifying actual key KPI anomalies; determining corresponding model evaluation metrics; evaluating the accuracy of the probabilistic statistical model using the model evaluation metrics; and determining the accuracy of the result regarding whether a key KPI is an anomaly based on the accuracy evaluation result of the probabilistic model. For anomaly detection, the evaluation metrics are more complex. The main reason is that anomaly detection datasets are generally imbalanced datasets, meaning there is more normal data than abnormal data. If we only look at the accuracy rate, we can imagine a scenario: if 90% of the data in the test set is normal, the accuracy rate can reach 90% when all predicted results are "normal". However, we are more concerned with anomalies. Therefore, the following method can be used to determine whether the anomaly detection result is correct; T / F represents whether the prediction and the actual result "match", P / N represents whether the prediction is "positive / negative", where P indicates a positive prediction and N indicates an anomaly prediction, as shown in the table below:

[0097]

[0098] From this, we can derive the following formula:

[0099]

[0100]

[0101]

[0102] R: Recall, significance (TPR) – the higher the value, the better the performance; P: Precision, representing the proportion of samples predicted as anomalous that are actually anomalous out of the total number of predicted anomalous samples – the higher the value, the better the performance.

[0103] The higher the weighted harmonic mean of F1:P and R, the better the performance. Therefore, anomaly detection focuses more on recall and precision metrics. The average recall and average precision of anomaly detection with N Kpi are calculated as follows:

[0104]

[0105]

[0106] At the same time, performing anomaly detection only on key KPIs can greatly improve detection efficiency.

[0107] For example, using actual weekly KPI time-series data, such as from March 26, 2022 to April 1, 2022, the following precision indicators can be obtained by calculating according to the above metrics:

[0108]

[0109]

[0110] This reflects the accuracy of the above probabilistic statistical model in detecting anomalies in key KPIs.

[0111] The following provides possible implementation methods for calculating the correlation between any two different KPIs and identifying related KPIs.

[0112] The correlation between any two distinct KPIs is calculated using a normalized cross-correlation method. KPIs with a correlation score greater than a set threshold are considered correlated. The sampled value x at time i of KPI x is then used. i The sampled value y of KPIy at time i i The correlation between KPIs x and y is calculated using the following formula:

[0113]

[0114] in:

[0115]

[0116]

[0117]

[0118]

[0119] The above 'n' refers to the number of KPIs collected from the start of data collection to time i. It refers to the average value of the sampled KPIx from the start of data collection to time i. It refers to the average value of the sampled KPIy from the start of data collection to time i, s x This refers to the variance of the KPIx sample values, s y This refers to the variance of the KPIy sample values.

[0120] For example, the correlation between wireless access success rate and RRC connection establishment success rate is 0.97. Figure 6 As shown.

[0121] The following provides a possible implementation method for determining the contribution of the counter factor to KPIs.

[0122] In one or more possible embodiments, determining the contribution of each counter to the KPI includes: taking each KPI at multiple times as a vector y, taking each counter that affects the KPI at the corresponding time as a vector x, and obtaining multiple discrete points; using partial least squares to perform linear fitting on the multiple discrete points to obtain vector k in the linear regression curve y = kx, and determining the vector k as the contribution of the counter to the KPI.

[0123] Based on the KPI calculation formula and prior knowledge of the relevant counters, partial least squares (PLS) is used to calculate the contribution of KPIs to the influencing factor counters. For example, the formula for calculating the success rate of RRC connection establishment is:

[0124] RRC.SuccConnEstab / RRC.AttConnEstab*100%

[0125] The calculation formula yields two counters: RRC.SuccConnEstab refers to the number of successful RRC connection establishments, and RRC.AttConnEstab refers to the ratio of RRC connection establishment requests to the total number of requests. There are four prior knowledge counters related to the RRC connection establishment success rate: RRC.FailConnEstab (acceptance failure), RRC.FailConnEstab.FailInRadioInterfaceProcedure.NoResponse (air interface timer timeout), RRC.FailConnEstab.Reject (cell rejection), and RRC.FailConnEstab.Others (other reasons). The calculation process involves abstracting the kpi data into a vector y, abstracting the counters related to that kpi into individual independent variables x, and using traditional partial least squares to train and obtain the k vectors within y = kx. Each k vector records the contribution of each x.

[0126] The following provides possible implementation methods for determining the importance of key KPIs.

[0127] The primary importance evaluation indicators include one or more of the following: whether it is a key KPI, whether it is an abnormal KPI, and the initial weights assigned to different non-key KPIs. For example, when evaluating the importance of key KPIs, the following formula is used for calculation:

[0128] IMT i =α×isKey(i)+β×isAnomaly(i)

[0129] Here, `isKey(i)` indicates whether a key KPI is a critical KPI, with a value of 0.7 for critical KPIs and 0.3 for non-critical KPIs. `isAnomaly(i)` indicates whether a key KPI has an outlier; if it does, it's an outlier with a value of 0.8, otherwise it's 0.2. α is 0.6 and β is 0.4. For example, a critical outlier KPI would have an IMT value of 0.6 * 0.7 + 0.4 * 0.8 = 0.74. When evaluating the importance of non-critical KPIs, it's necessary to check for correlation with critical KPIs. If a correlation exists, the non-critical KPI's anomaly is determined based on the associated critical KPI. Therefore, the initial weight set for a non-critical KPI, which is its importance, can be either 0.6 * 0.3 + 0.4 * 0.8 = 0.5 or 0.6 * 0.3 + 0.4 * 0.2 = 0.26.

[0130] According to the KPI-based root cause localization method provided in this application, faults can be located quickly and efficiently, the specific KPIs that caused the faults and their root causes can be identified, and fault recovery time can be reduced, improving operation and maintenance efficiency and user experience.

[0131] Based on the same inventive concept, in a second aspect, this application also provides a KPI-based root cause localization device, such as... Figure 7 As shown, it includes:

[0132] The correlation determination module 701 is used to calculate the correlation between any two different KPIs based on the sampled values ​​of two different KPIs at the same time for all KPIs of each target cell, and to determine the KPIs that are related.

[0133] The contribution determination module 702 is used to determine the contribution of each counter to the KPI for each target cell based on the sampled values ​​of each KPI at different times and the counters of each influencing factor affecting the KPI at the corresponding times.

[0134] The importance determination module 703 is used to determine the importance of each KPI based on the different first importance evaluation indicators of each KPI in each target cell;

[0135] The influence determination module 704 is used to determine the fault influence of any target cell based on the importance of any KPI, the correlation degree of other KPIs related to this KPI, and the contribution of each counter that affects this KPI.

[0136] The fault determination module 705 is used to determine the root cause of a fault based on the fault impact of the KPIs of each target cell.

[0137] Based on the same inventive concept, this application also provides a KPI-based root cause localization device in a third aspect. The device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform any of the KPI-based root cause localization methods provided in the embodiments of the first aspect.

[0138] like Figure 8 As shown, the device includes a processor 801, a memory 802, a communication interface 803, and a bus 804. The processor 801, memory 802, and communication interface 803 are interconnected via the bus 804.

[0139] The processor 801 is used to read instructions from the memory 802 and execute them, so that the at least one processor can execute a method for locking network type and frequency band provided in the above embodiments.

[0140] The memory 802 is used to store various instructions and programs for the inter-system measurement configuration anomaly handling method provided in the above embodiments.

[0141] The 804 bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be divided into address buses, data buses, control buses, etc. For ease of representation, Figure 8 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0142] The processor 801 can be a central processing unit (CPU), a network processor (NP), a graphics processing unit (GPU), or any combination of CPU, NP, and GPU. It can also be a hardware chip. The aforementioned hardware chip can be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The aforementioned PLD can be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0143] Based on the same inventive concept, in a fourth aspect, this application also provides a storage medium storing a computer program for causing a computer to perform any of the KPI-based root cause localization methods provided in the embodiments of the first aspect.

[0144] like Figure 9 As shown, the memory may include a readable medium in the form of volatile memory, such as random access memory (RAM) 1321 and / or cache memory 1322, and may further include read-only memory (ROM) 1323.

[0145] The memory may also include a program / utility 1325 having a set (at least one) of program modules 1324, including but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment. Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0146] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0147] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0148] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0149] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A KPI-based root cause localization method, characterized in that, include: Obtain at least one key performance indicator (KPI) for each of the multiple cells; When it is determined that any acquired key KPI has a preset threshold, the key KPI is compared with the corresponding preset threshold to determine whether the key KPI is abnormal; when it is determined that any acquired key KPI does not have a preset threshold, a probabilistic statistical model is used to predict whether the key KPI is abnormal; target cells with abnormal key KPIs are screened out; for all KPIs of each target cell, the correlation between each pair of different KPIs is calculated based on the sampled values ​​of two different KPIs in the same time period to determine the KPIs that are related. For each target cell, based on the sampled values ​​of each KPI at multiple different times and the counters of each influencing factor affecting the KPI at the corresponding times, the contribution of each counter to the KPI is determined. The importance of each KPI is determined based on the different first-importance evaluation indicators for each KPI in each target community; For any target cell, the fault impact of a KPI is determined by comprehensively considering the importance of any KPI, the correlation of other KPIs related to that KPI, and the contribution of each counter that affects that KPI. The KPIs of each target cell are sorted according to their fault impact, and the top n KPIs with the largest fault impact are determined. The contribution of the counters of each of the top n KPIs is sorted, and the top m counters with the largest contribution to each KPI are determined as the root causes of the fault, where m and n are positive integers.

2. The method according to claim 1, characterized in that, Also includes: The importance of each counter is determined based on the different second importance evaluation indicators of each counter in each target cell; For any target cell, the fault impact of a counter is determined based on the importance of any counter and the contribution of that counter to each KPI.

3. The method according to claim 2, characterized in that, Map each KPI and counter to nodes, establish connections between nodes corresponding to two related KPIs, and between nodes corresponding to KPIs and counters with influence relationships. Determine the failure impact of each KPI and the failure impact of each counter, specifically including: For any given node, the node's own influence is determined based on its importance and the first weighting factor. The influence of other nodes connected to the node on the node is determined based on the correlation / contribution of each of the other nodes connected to the node and the second weighting factor. The failure impact of a node is determined based on the node's influence and the influence exerted on the node by other nodes connected to it. Wherein, the first weighting factor is the reciprocal of the total number of nodes, and the second weighting factor is the failure impact of other nodes connected to this node.

4. The method according to claim 1, characterized in that, Also includes: Map each KPI and counter to a node; Connect the nodes corresponding to two related KPIs through a first edge, where the weight of the first edge is the degree of correlation between the two KPIs. Connect the nodes corresponding to the KPIs and counters that have an influence relationship through a second edge to obtain the KPI association propagation graph. The weight of the second edge is the contribution of the counter to the KPI.

5. The method according to claim 1, characterized in that, Using probabilistic statistical models to predict whether this key KPI is abnormal, including: Input all sampled values ​​of any key KPI that does not have a preset threshold into the probability and statistics model; The probabilistic statistical model is used to aggregate all the sampled values ​​into two categories, and the anomaly category is determined based on the number of sampled values ​​in each category. Determine whether the key KPI is abnormal based on the category to which the current sampled value of the key KPI belongs.

6. The method according to claim 1 or 5, characterized in that, Also includes: The accuracy of the probability and statistics model is evaluated, and the accuracy of the results of the accuracy evaluation is determined based on the results of the probability and statistics model evaluation to determine whether the key KPIs are abnormal.

7. The method according to claim 1, characterized in that, Determine the contribution of each counter to this KPI, including: By taking each KPI at multiple times as a vector y and taking each counter that affects the KPI at the corresponding time as a vector x, multiple discrete points are obtained. The partial least squares method is used to perform linear fitting on the multiple discrete points to obtain the vector k in the linear regression curve y=kx, and the vector k is determined as the contribution of the counter to the KPI.

8. The method according to claim 1, characterized in that, The first importance evaluation index includes any one or more of the following: Whether it is a key KPI, whether it is an abnormal KPI, and the first initial weight set for different non-key KPIs.

9. The method according to claim 2, characterized in that, The second importance evaluation index is a second initial weight set for different counters, or the weight of each counter is calculated based on whether the KPIs related to each counter are abnormal.

10. The method according to claim 1, characterized in that, Calculate the correlation between any two distinct KPIs to identify the correlated KPIs, specifically including: The correlation between any two different KPIs is calculated using the normalized cross-correlation method, and two KPIs with a correlation greater than a set threshold are identified as related KPIs.

11. The method according to claim 1, characterized in that, The key KPIs include any one or more of the following: Radio access success rate, UE context establishment success rate, Radio Resource Control (RRC) connection establishment success rate, and Radio Resource Control (RRC) reconstruction success rate.

12. A KPI-based root cause localization device, characterized in that, include: The correlation determination module is used to obtain at least one key performance indicator (KPI) corresponding to each of multiple cells. When it is determined that any acquired key KPI has a preset threshold, the key KPI is compared with the corresponding preset threshold to determine whether the key KPI is abnormal; when it is determined that any acquired key KPI does not have a preset threshold, a probabilistic statistical model is used to predict whether the key KPI is abnormal; target cells with abnormal key KPIs are screened out; for all KPIs of each target cell, the correlation between each pair of different KPIs is calculated based on the sampled values ​​of two different KPIs at the same time, and the KPIs with correlation are determined. The contribution determination module is used to determine the contribution of each counter to the KPI for each target cell based on the sampled values ​​of each KPI at different times and the counters of each influencing factor affecting the KPI at the corresponding times. The importance determination module is used to determine the importance of each KPI based on the different first importance evaluation indicators for each KPI in each target cell; The impact determination module is used to comprehensively determine the fault impact of any target cell based on the importance of any KPI, the correlation degree of other KPIs related to this KPI, and the contribution of each counter affecting this KPI. The fault determination module is used to sort the KPIs of each target cell according to their fault impact, and determine the top n KPIs with the largest fault impact; sort the contribution of the counters of each KPI among the top n KPIs, and determine the top m counters with the largest contribution to each KPI as the root cause of the fault, where m and n are positive integers.

13. A KPI-based root cause localization device, characterized in that, The device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the KPI-based root cause localization method as described in any one of claims 1-11.

14. A storage medium, characterized in that, The storage medium stores a computer program for causing a computer to execute the KPI-based root cause localization method as described in any one of claims 1-11.

Images