WO2022028120A1

WO2022028120A1 - Indicator detection model acquisition method and apparatus, fault locating method and apparatus, and device and storage medium

Info

Publication number: WO2022028120A1
Application number: PCT/CN2021/100381
Authority: WO
Inventors: 冯媛
Original assignee: 中兴通讯股份有限公司
Priority date: 2020-08-06
Filing date: 2021-06-16
Publication date: 2022-02-10
Also published as: CN114095965A

Abstract

Provided are an indicator detection model acquisition method and apparatus, a fault locating method and apparatus, and a device and a storage medium. The indicator detection model acquisition method in the present application comprises: according to a preset KPI system, acquiring KPIs of leaf nodes in a mobile communication network, wherein the KPI system comprises at least two types of KPIs; after the KPIs of the leaf nodes are respectively aggregated by type, extracting a feature parameter of each type of KPI after the aggregation of the leaf nodes; and respectively taking the feature parameter of the same type of KPI of the leaf nodes as an input of a preset model to perform model training, so as to obtain an indicator detection model corresponding to each type of KPI.

Description

Index detection model acquisition and fault location method, device, equipment and storage medium

cross reference

This application is based on the Chinese patent application with the application number "202010783870.6" and the application date is August 6, 2020, and claims the priority of the Chinese patent application. The entire content of the Chinese patent application is hereby incorporated by reference. Apply.

technical field

Embodiments of the present application relate to the field of failure analysis of semiconductor devices, and in particular, to a method, apparatus, device, and storage medium for acquiring an index detection model and locating faults.

Background technique

In the field of IT operation and maintenance of mobile operators, multi-dimensional and multi-index fault root cause node location has always been the core issue of intelligent operation and maintenance (AIOPS). When the key performance indicators of a certain network element node are abnormal, the operation and maintenance personnel need to quickly and accurately locate the cause of the failure, so as to carry out repair and stop loss work. However, due to the complexity of mobile networking and the diversity of indicators, it is difficult and time-consuming to troubleshoot root cause nodes. The operation and maintenance personnel hope that the results of root cause nodes given by the system are as concise as possible. That is, the cause of the fault is covered with the least dimension and attribute values, so that the operation and maintenance personnel can quickly verify and fix it.

At present, mobile operators have begun to monitor key KPI performance indicators in the network, and use KPI thresholds to determine whether the indicators are abnormal. This method will lead to misjudgment of indicators that fall within the gray interval of the threshold edge, which will lead to a large number of false alarms and false alarms, and it is difficult to truly reflect the network quality; and a large number of alarm information will interfere with the rapid and accurate judgment of operation and maintenance personnel. Failure and analysis of the cause result in low efficiency and high cost of fault repair, which in turn affects network performance.

SUMMARY OF THE INVENTION

An embodiment of the present application provides a method for obtaining an index detection model for a mobile communication network, including: obtaining key KPI indexes of each leaf node in a mobile communication network according to a preset key KPI index system, where the key KPI index system includes: At least two types of key KPI indicators; after aggregating the key KPI indicators of each leaf node by category, extract the characteristic parameters of each type of key KPI indicators aggregated by each leaf node; The characteristic parameters of the model are used as the input of the preset model for model training, and the index detection model corresponding to each type of key KPI index is obtained.

The embodiment of the present application also provides a fault location method for a mobile communication network, including: using the index detection model obtained by the above method for obtaining an index detection model for a mobile communication network, to each leaf node in the mobile communication network Perform index detection; for the target leaf node whose detection result is that the key KPI index is abnormal, it is determined whether the target leaf node is a fault critical node based on the historical status of the target leaf node.

The embodiment of the present application also provides a device for obtaining an index detection model for a mobile communication network, including: an obtaining module, configured to obtain the key KPI index of each leaf node in the mobile communication network according to a preset key KPI index system, The key KPI indicator system includes at least two types of key KPI indicators; the training module is used to aggregate the key KPI indicators of each leaf node by category, and then extract the characteristic parameters of each type of key KPI indicators aggregated by each leaf node, And the feature parameters of the same type of key KPI indicators of each leaf node are used as the input of the preset model for model training, and the indicator detection model corresponding to each type of key KPI indicators is obtained.

The embodiment of the present application further provides a fault location device for a mobile communication network, including: an abnormality detection module, configured to use the index detection model obtained by the above-mentioned method for obtaining an index detection model for a mobile communication network to Each leaf node in the network performs index detection; the fault location module is used to determine whether the target leaf node is a fault critical node based on the historical status of the target leaf node for which the detection result is an abnormal key KPI index.

Embodiments of the present application also provide an electronic device, including a processor, a memory, and a communication bus connecting the processor and the memory; the memory stores a first computer program, and the first computer program can be executed by the processor to realize the above-mentioned functions and/or, the memory stores a second computer program, and the second computer program can be executed by the processor to implement the steps of the above method for locating faults in the mobile communication network.

Embodiments of the present application further provide a computer-readable storage medium, where the computer-readable storage medium stores a first computer program, and the first computer program can be executed by a processor to implement the above index detection model for a mobile communication network and/or, the computer-readable storage medium stores a second computer program, and the second computer program can be executed by the processor to implement the steps of the above method for locating faults in a mobile communication network.

Other features and corresponding beneficial effects of the present application are described in later parts of the specification, and it should be understood that at least some of the beneficial effects will become apparent from the description in the specification of the present application.

Description of drawings

1 is a schematic flowchart of a method for obtaining an index detection model for a mobile communication network provided by Embodiment 1 of the present application;

2 is a schematic diagram of a key KPI indicator system provided in Embodiment 1 of the present application;

3 is a schematic flowchart of a method for locating a fault in a mobile communication network according to Embodiment 1 of the present application;

4 is a schematic diagram of a process of locating a root cause node of a fault according to Embodiment 1 of the present application;

5 is a schematic structural diagram of an apparatus for obtaining an index detection model for a mobile communication network according to Embodiment 2 of the present application;

FIG. 6 is a schematic structural diagram of a fault locating apparatus for a mobile communication network according to Embodiment 2 of the present application;

FIG. 7 is a schematic diagram of an index detection model acquisition and fault location process provided in Embodiment 2 of the present application;

FIG. 8 is a schematic diagram of probe deployment according to Embodiment 2 of the present application;

FIG. 9 is a schematic diagram of topology networking according to Embodiment 2 of the present application;

FIG. 10-1 is a schematic diagram of a characteristic parameter matrix of key KPI indicators of a cell according to Embodiment 2 of the present application;

10-2 is a schematic diagram of a characteristic parameter matrix of key KPI indicators of a network element according to Embodiment 2 of the present application;

FIG. 11 is a schematic diagram of a link provided in Embodiment 2 of the present application;

FIG. 12 is a schematic structural diagram of an electronic device according to Embodiment 3 of the present application.

detailed description

In order to make the purpose, technical solutions and advantages of the present application more clearly understood, the embodiments of the present application will be further described in detail below through specific implementation manners in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application.

The embodiments of the present application mainly provide a method, device, equipment, and storage medium for obtaining an index detection model and locating faults, which solve the problem that in the related art, the accuracy of fault locating in mobile communication networks is poor, resulting in low fault repair efficiency and high cost, which in turn affects the network. performance issue.

The method, device, device, and storage medium for acquiring an indicator detection model and locating faults provided by the embodiments of the present application collect key KPI indicators and extract characteristic parameters for each leaf node in a mobile communication network through a preset key KPI indicator system. , and the extracted feature parameters are used as the input of the preset model to train the model to obtain the index detection model, and then the index detection model can be used to detect abnormal indexes of each leaf node in the mobile communication network, and can be combined with the leaves with abnormal indexes. The historical status of the node can accurately determine whether the leaf node is a key fault node. Compared with the threshold-based indicator detection and the fault location based on the experience of the operation and maintenance personnel in the related technology, it no longer relies on the threshold threshold to divide the abnormality and the experience of the operation and maintenance personnel. , which can effectively avoid false alarms and missed alarms, accurately detect leaf nodes with abnormal indicators, and determine whether the node is a key fault node based on the historical status of the leaf node, which greatly reduces the alarm range and improves the accuracy of fault location. This can improve the efficiency of subsequent fault repair, reduce repair costs, and improve network performance.

Example 1:

The specific embodiments of the present application will be described in further detail below with reference to the accompanying drawings and implementation examples.

The KPI threshold is used in related technologies to determine the abnormality of the indicators, which leads to misjudgment of the indicators falling within the gray interval of the threshold edge, which leads to a large number of false alarms and false alarms, and it is difficult to truly reflect the network quality; and a large number of alarm information It will interfere with the operation and maintenance personnel to quickly and accurately determine the fault and analyze the cause, resulting in low fault repair efficiency and high cost, which will affect the network performance. This embodiment first provides a method for obtaining an index detection model that can be used in a mobile communication network. , the index detection model obtained by this method can accurately detect whether the key KPI indicators of each leaf node in the mobile communication network are abnormal, and no longer rely on the threshold threshold to divide the abnormality and the experience of operation and maintenance personnel, and can effectively Avoid false alarms and missed alarms.

Referring to FIG. 1 , the method for obtaining an index detection model for a mobile communication network provided by this embodiment may include, but is not limited to:

S101: Acquire key KPI indicators of each leaf node in the mobile communication network according to a preset key KPI indicator system, where the key KPI indicator system includes at least two types of key KPI indicators.

It should be understood that the key KPI indicator types included in the key KPI indicator system in this embodiment may be flexibly set according to specific application scenarios and requirements. For example, for a VOLTE (Voice over Long-Term Evolution, long-term evolution voice bearer) network, the key KPI indicator system is shown in Figure 2, which can include but not limited to at least two of the following types of key KPI indicators. Flexible selection, and other types of key KPI indicators can also be expanded or replaced according to needs:

Initial registration success rate;

Re-registration success rate;

VoLTE network connection rate;

V2V call establishment duration;

VoLET call drop rate;

eSRVCC (Enhanced Single Radio Voice Call Continuity, enhanced single wireless voice call continuity) handover success rate;

eSRVCC switching average delay;

The proportion of VoLET MOS3.0;

Upstream packet loss rate.

It should be understood that, in this embodiment, when the key KPI indicators are collected for each leaf node in the mobile communication network, the key KPI indicators in the same time period of each leaf node may be collected. Optionally, in some application examples, key KPI indicators of each leaf node in different time periods may also be collected.

It should be understood that, in this embodiment, when the key KPI indicators are collected for each leaf node in the mobile communication network, the types of the key KPI indicators collected for each leaf node may be the same. Optionally, in some application examples, there may also be types of key KPI indicators collected for some leaf nodes, which are different from the types of key KPI indicators collected by other leaf nodes; the specific types can be flexibly set according to requirements.

It should be understood that, in this embodiment, when the key KPI indicators are collected for each leaf node in the mobile communication network, the collection may be performed periodically, for example, the collection period (that is, the collection granularity) may be set in hours, days, or months. ; Of course, optionally, in some application examples, periodic collection may not be used for collection.

Optionally, in some embodiments, registration signaling, call signaling, call drop signaling, eSRVCC handover signaling, voice quality signaling, etc. may be collected at corresponding locations in the mobile communication network, but not limited to. The key KPI indicators of each leaf node are obtained, and the collected signaling may include but not limited to time information, user information, behavior information, location information, and key KPI indicators.

S102: After aggregating the key KPI indicators of each leaf node by category, extract characteristic parameters of each category of key KPI indicators aggregated by each leaf node.

For example, for a leaf node, the initial registration success rate, re-registration success rate, VoLTE network connection rate, V2V call establishment duration, VoLET call drop rate, eSRVCC handover success rate, eSRVCC handover average delay, VoLET MOS3 .0 ratio, uplink packet loss rate. Then, the above various key KPI indicators of the leaf node are aggregated to obtain a set of various KPI indicators above each leaf node.

Then, for the leaf node, the characteristic parameters of various key KPI indicators are extracted. The feature parameters extracted in this embodiment may include, but are not limited to, at least two of the following, which may be flexibly selected according to requirements, and may also be flexibly expanded or replaced according to requirements:

The mean value of key KPI indicators, the standard deviation of key KPI indicators, the minimum value of key KPI indicators, the maximum value of key KPI indicators, the quarter point of key KPI indicators, the median value of key KPI indicators, the The third-quarter point, the standard deviation mean of key KPI indicators, the variance mean of key KPI indicators, the Chebyshev statistical characteristics of key KPI indicators, the total variation of key KPI indicators, and the coefficient of variation of key KPI indicators.

S103: Use the characteristic parameters of the same type of key KPI indicators of each leaf node as the input of the preset model to perform model training, and obtain an indicator detection model corresponding to each type of key KPI indicators.

For example, assuming that there are 80 leaf nodes in the current mobile network, and the nine types of key KPI indicators in the above example are collected for each leaf node, the characteristic parameters of each of the same type of key KPI indicators for the 80 leaf nodes are used as pre- The input of the model is set for model training, so as to obtain the index detection model of the nine types of key KPI indicators in the above example. Through this indicator detection model, it is possible to detect whether the key KPI indicators of the corresponding type on the leaf nodes are abnormal.

It should be understood that the preset model in this embodiment may be flexibly selected according to specific application scenarios. For example, in an example, the preset model may be, but not limited to, a random forest model. In this example, the characteristic parameters of each type of key KPI indicators after the aggregation of each leaf node are extracted include:

For each leaf node, extract the characteristic parameters of the aggregated key KPI indicators of each type of key KPI indicators that fall within the first time window, and the characteristic parameters of the key KPI indicators that fall within the second time window, where the second The time window is greater than the first time window, for example, the first time window can be 7 days, the second time window can be 30 days, or the first time window is 4 hours, the second time window is 40 hours, etc., that is, the first time window The size of the window and the second time window can be flexibly set according to specific application requirements.

In this example, the feature parameters of the same type of key KPI indicators of each leaf node are used as the input of the preset model for model training, including:

The feature parameters of the same type of key KPI indicators of each leaf node are converted into time series as the input of the random forest model, and the random forest model is trained to obtain the corresponding indicator detection model.

This embodiment also provides a fault location method for a mobile communication network, as shown in FIG. 3 , which includes but is not limited to:

S301: Using the index detection model obtained by the above-described method for obtaining an index detection model for a mobile communication network, perform index detection on each leaf node in the mobile communication network.

Wherein, through the index detection model, the index detection of each leaf node in the mobile communication network may include but is not limited to:

The time series sequence of each leaf node is predicted through the index detection model, and when the difference between the predicted value and the actual value (that is, the actual value of the time series sequence of the leaf node) is greater than the set threshold, it is determined that the target leaf node corresponds to the index detection The key KPI indicators of the model are abnormal.

In this embodiment, when indicators are detected for each leaf node, it is possible to detect whether the above-mentioned various key KPI indicators of the leaf node are normal, so as to determine whether various key KPI indicators are normal, and when there is a certain type of KPI indicators When abnormal, the leaf node can be considered as the target leaf node.

S302: For the target leaf node whose detection result is that the key KPI indicator is abnormal, determine whether the target leaf node is a fault critical node in combination with the historical state of the target leaf node. For example, this step may include, but is not limited to:

When the historical status of the target leaf node is that the key KPI indicators are normal, it is determined that the target leaf node is not a fault critical node;

When the historical state of the target leaf node is abnormal for the key KPI indicators, it is determined that the target leaf node is the key fault node;

In this embodiment, the historical state of the leaf node includes the historical state of the key KPI indicators determined to be abnormal in the above S301; and the retroactive historical time value can be flexibly set according to specific application scenarios, for example, it can be set to the previous several hours, several days Or a month or a few months and so on.

In this embodiment, in order to avoid the ripple effect of abnormal fluctuations in the complex networking, the fluctuation contribution of each target leaf node determined as the critical node of the failure can also be calculated, and according to the calculated result, the contribution of each target leaf node determined as the critical node of the failure The root cause node of the fault is determined in the target leaf node, thereby further improving the accuracy of fault location. In this embodiment, the fluctuation contribution degree is a value that reflects the degree of influence on other leaf nodes connected to the target leaf node when the target leaf node fails.

When there are multiple target leaf nodes determined to be the critical fault nodes in S302, the location process of the fault root cause node in the above example is shown in FIG. 4, including but not limited to:

S401: Acquire the fluctuation contribution degree of each target leaf node determined as the fault critical node.

S402: Determine the target leaf node with the largest fluctuation contribution as the fault root cause node.

Optionally, in some examples in this embodiment, before the index detection is performed on each leaf node in the mobile communication network by using the index detection model, the index detection model may further include but not limited to:

According to the obtained user plane signaling and control plane signaling, the topology structure of the mobile communication network is self-learned, so as to obtain the topology structure of the mobile communication network, and each network entity in the obtained topology structure is used as a leaf node .

Compared with the traditional threshold-based indicator detection and fault cause location relying on the experience of operation and maintenance personnel, the method for acquiring the indicator detection model and the method for locating faults provided in this embodiment adopts the learning method based on the historical characteristics (ie, characteristic parameters) of multi-dimensional KPI indicators. , identify key KPI indicators; and can analyze the root cause node of the fault in combination with the topology of the mobile communication network on site; in the process of abnormal identification, compare the key KPI indicators of the leaf node with its own historical data without relying on the threshold The threshold is divided into anomalies, which can effectively avoid false alarms and missed alarms, and can detect obvious anomalies of key KPI indicators and anomalies with deterioration trends; in the part of fault root cause node location, combined with the results of KPI anomaly detection and the historical status of the node , to determine whether the node is the key node of the fault, and combined with the topology of the network, the root cause node of the fault can be determined by calculating the maximum fluctuation contribution, which solves the problem of the ripple effect of the index fluctuation, thereby filtering out the error caused by the occasional index fluctuation. alarm, greatly reducing the scope of the alarm. In this way, the most comprehensive fault root cause node can be realized with the combination of as few dimensions and their attribute values as possible.

Embodiment 2:

This embodiment also provides a device for obtaining an index detection model for a mobile communication network, as shown in FIG. 5 , which includes but is not limited to:

The acquisition module 501 is configured to acquire the key KPI indicators of each leaf node in the mobile communication network according to the preset key KPI indicator system, the key KPI indicator system includes at least two types of key KPI indicators, and the specific acquisition process is shown in the above embodiment. , and will not be repeated here.

The training module 502 is used to aggregate the obtained key KPI indicators of each leaf node by category, extract the characteristic parameters of each type of key KPI indicators after the aggregation of each leaf node, and separate the same type of key KPI indicators of each leaf node respectively. The characteristic parameters of KPI indicators are used as the input of the preset model for model training, and the indicator detection model corresponding to each type of key KPI indicators is obtained. The specific training process is shown in the above-mentioned embodiment, and details are not described herein again.

The functions of the acquisition module 501 and the training module 502 in this embodiment may be implemented by, but not limited to, a processor.

This embodiment also provides a fault location device for a mobile communication network, as shown in FIG. 6 , which includes but is not limited to:

The anomaly detection module 601 is configured to perform index detection on each leaf node in the mobile communication network by using the index detection model obtained by the above-described index detection model acquisition method for the mobile communication network. The specific detection process is shown in the above examples, and details are not repeated here.

The fault location module 602 is used to determine whether the target leaf node is a fault key node for the target leaf node for which the detection result is that the key KPI index is abnormal, combined with the historical state of the target leaf node; the specific determination process is shown in the above embodiment. It is not repeated here.

In some examples of this embodiment, the fault locating apparatus further includes a topology networking self-learning module 603, configured to perform self-learning on the topology structure of the mobile communication network according to the acquired user plane signaling and control plane signaling, thereby The topology structure of the mobile communication network is acquired, and each network entity in the acquired topology structure is used as a leaf node. In this way, when there is a networking change in the network, the latest networking structure can be obtained in real time without relying on a third-party data interface.

For ease of understanding, this embodiment is further described below with an application example.

In this application example, an application scenario of a mobile communication network based on a VoLTE network is used as an example for description. In this application example, the process of acquiring an indicator detection model for a mobile communication network and locating a fault is shown in Figure 7, including but not limited to network indicator collection S701, raw data cleaning S702, multi-dimensional indicator aggregation S703, and key KPI indicators Feature parameter extraction S704, model training S705, index anomaly identification S706, and fault root cause location S707 to complete the process of anomaly identification and root cause location.

An example of the network indicator collection method is shown in Figure 8. Hard probes can be deployed on the Mw port between the I/M-CSCF and SBC/P-CSCF to collect registration signaling, call signaling, and dropped calls. Class signaling; the Rx port deployed on the P/SBC and PCRF collects call drop signaling; the Sv port deployed between the MME and eMSC network elements collects eSRVCC handover signaling; the s1-u deployed between the cell and the SGW network element The port collects voice quality signaling. The collected signaling includes time information, user information, behavior information, location information, and key KPI indicators. The key KPI system in this application example is shown in Figure 2.

An example of raw data cleaning may include: cleaning, stripping, and normalizing valid information in each collected signaling information, and standardizing and normalizing the data. At the same time, it is also necessary to combine the VoLTE business logic to clean the topology structure of the existing network in real time. For example, according to the obtained user plane instructions and control plane instructions, the on-site mobile communication network topology structure is self-learned. An example of self-learning The schematic diagram of the topology networking is shown in Figure 9, and the cell Ci and each network element on each core network side in the figure are used as leaf nodes.

The multi-dimensional index aggregation is to aggregate the indicators of the cell dimension and each network element dimension for the cleaned data, and collect statistics at a set time granularity (for example, hours).

In this application example, the feature parameter extraction of key KPI indicators includes the mean value of key KPI indicators, the standard deviation of key KPI indicators, the minimum value of key KPI indicators, the maximum value of key KPI indicators, and a quarter of key KPI indicators Quantile, median value of key KPI indicators, quartiles of key KPI indicators, standard deviation mean of key KPI indicators, mean variance of key KPI indicators, Chebyshev statistical characteristics of key KPI indicators, key KPIs The total variation of indicators, the coefficient of variation of key KPI indicators, and the characteristic parameters are converted into matrices, as shown in Figure 10-1 and Figure 10-2. The first column of the matrix shown in Figure 10-1 is each cell as a leaf node, that is, the characteristic parameters of a certain type of key KPI indicators of the network entity on the wireless network side. For the kpi01t1 to ci1 corresponding to the first row kpi01tn corresponds to the mean value of key KPI indicators, the standard deviation of key KPI indicators, the minimum value of key KPI indicators, the maximum value of key KPI indicators, and the quarter quantile of key KPI indicators of a certain type of key KPI indicators of ci1 point, the median value of key KPI indicators, the third-quarter point of key KPI indicators, the standard deviation mean of key KPI indicators, the variance mean of key KPI indicators, the Chebyshev statistical characteristics of key KPI indicators, the average value of key KPI indicators Total variation, coefficient of variation of key KPI indicators. The first column of the matrix shown in Figure 10-2 is each network element serving as a leaf node, that is, a network entity on the core network side.

The model training in this application example includes converting the feature parameters shown in Figure 10-1 and Figure 10-2 into time series, which are used as the input of the random forest model for model training to obtain an indicator detection model.

In the indicator anomaly identification in this application example, the obtained indicator detection model can be used to predict the key KPI time series of each leaf node. When the difference between the predicted result and the actual result is greater than the preset threshold, the indicator is determined to be an abnormal indicator. Determine the corresponding leaf node as the target leaf node. It is assumed here that the key KPI indicators of 30 cells are abnormal, and the key KPI indicators of 7 network elements are abnormal.

The fault root cause location in this application example may include: acquiring the historical status information of the target leaf node (that is, the above-mentioned 30 cells and 7 network elements), if the historical status is normal, the target leaf node is determined to be a key fault node; If the historical state is abnormal, the target leaf node is determined not to be the critical node of the failure. Suppose that after the above determination, it is found that 10 cells among the above 30 cells are fault critical nodes, and 3 network elements among the above 7 network elements are fault critical nodes.

Combined with the topology structure of the entire network, the fluctuation contribution of each key fault node is distributed and calculated. A schematic diagram of a link structure is shown in Figure 11. The uplink packet loss rate indicator wave of the MME numbered 14XX87 is abnormal, and its fluctuation contribution is 0.86 (for example, if the number of other leaf nodes connected to the MME is 100, the abnormal packet loss rate indicator of the MME will affect the 86, the dynamic contribution of the MME can be determined to be 0.86; of course, other calculation methods can also be used, as long as the dynamic contribution defined in this embodiment can be accurately obtained), which is the maximum fluctuation contribution in the group , so it is determined that the root is the uplink packet loss rate indicator of the MME numbered 14XX87.

Embodiment three:

This embodiment also provides an electronic device, as shown in FIG. 12, including a processor 1201, a memory 1202, and a communication bus 1203 connecting the processor 1201 and the memory 1202;

The memory 1202 stores a first computer program, and the first computer program can be executed by the processor 1201 to implement the steps of the method for obtaining an index detection model for a mobile communication network as shown in the above embodiments;

And/or, the memory 1202 stores a second computer program, and the second computer program can be executed by the processor 1201 to implement the steps of the mobile communication network fault location method shown in the above embodiments.

This embodiment also provides a computer-readable storage medium, where the computer-readable storage medium stores a first computer program, and the first computer program can be executed by a processor, so as to implement the mobile communication as shown in the above embodiments The steps in the method for obtaining the index detection model of the network;

And/or, the computer-readable storage medium stores a first computer program, and the first computer program can be executed by the processor to implement the steps in the mobile communication network fault location method shown in the above embodiments.

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

This embodiment also provides a first computer program (or referred to as first computer software), the first computer program can be distributed on a computer-readable medium and executed by a computable device, so as to realize the above-mentioned and in some cases, at least one of the steps shown or described may be performed in an order different from that described in the above embodiments.

This embodiment also provides a second computer program (or called second computer software), the second computer program can be distributed on a computer-readable medium and executed by a computer-readable device, so as to realize the above-mentioned mobile communication and, in some cases, at least one of the steps shown or described may be performed in an order different from that described in the above embodiments.

This embodiment also provides a computer program product, including a computer-readable device, on which any of the computer programs shown above are stored. In this embodiment, the computer-readable device may include the computer-readable storage medium as described above.

It can be seen that those skilled in the art should understand that all or some of the steps in the methods disclosed above, the functional modules/units in the system, and the device can be implemented as software (which can be implemented by computer program codes executable by a computing device). ), firmware, hardware, and their appropriate combination. In a hardware implementation, the division between functional modules/units mentioned in the above description does not necessarily correspond to the division of physical components; for example, one physical component may have multiple functions, or one function or step may be composed of several physical components Components execute cooperatively. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor or microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit .

In addition, communication media typically embodies computer readable instructions, data structures, computer program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and can include any information delivery, as is well known to those of ordinary skill in the art medium. Therefore, the present application is not limited to any particular combination of hardware and software.

The above content is a further detailed description of the embodiments of the present application in conjunction with specific implementations, and it cannot be considered that the specific implementation of the present application is limited to these descriptions. For those of ordinary skill in the technical field of the present application, without departing from the concept of the present application, some simple deductions or substitutions can be made, which should be regarded as belonging to the protection scope of the present application.

Claims

A method for obtaining an index detection model for a mobile communication network, comprising:

According to the preset key KPI index system, obtain the key KPI index of each leaf node in the mobile communication network, and the key KPI index system includes at least two types of key KPI indexes;

After aggregating the key KPI indicators of each leaf node by category, extract the characteristic parameters of each type of key KPI indicators aggregated by each leaf node;

The characteristic parameters of the same type of key KPI indicators of each leaf node are respectively used as the input of the preset model for model training, and the indicator detection model corresponding to each type of key KPI indicators is obtained.
The method for obtaining an index detection model for a mobile communication network according to claim 1, wherein the preset model is a random forest model, and the extracting characteristic parameters of each type of key KPI index aggregated by each leaf node include: :

For each leaf node, extract the characteristic parameters of the key KPI indicators that fall within the first time window for each type of key KPI indicators after aggregation, and the characteristic parameters of the key KPI indicators that fall within the second time window. The second time window is greater than the first time window;

The model training using the characteristic parameters of the same type of key KPI indicators of each leaf node as the input of the preset model respectively includes:

The characteristic parameters of the same type of key KPI indicators of each leaf node are respectively converted into time series as the input of the random forest model, and the random forest model is trained to obtain an indicator detection model.
The method for obtaining an index detection model for a mobile communication network according to claim 1 or 2, wherein the characteristic parameters include at least two of the following:

The mean value of key KPI indicators, the standard deviation of key KPI indicators, the minimum value of key KPI indicators, the maximum value of key KPI indicators, the quarter point of key KPI indicators, the median value of key KPI indicators, the The third-quarter point, the standard deviation mean of key KPI indicators, the variance mean of key KPI indicators, the Chebyshev statistical characteristics of key KPI indicators, the total variation of key KPI indicators, and the coefficient of variation of key KPI indicators.
The method for obtaining an index detection model for a mobile communication network according to any one of claims 1 to 3, wherein the key KPI index system includes at least two types of the following key KPI indexes:

Initial registration success rate, re-registration success rate, VoLTE network connection rate over long-term evolution voice, V2V call setup duration, VoLET call drop rate, enhanced single wireless voice call continuity eSRVCC handover success rate, eSRVCC handover average delay, VoLET MOS3.0 proportion, uplink packet loss rate.
A fault location method for a mobile communication network, comprising:

Using the index detection model obtained by the method for obtaining an index detection model for a mobile communication network according to any one of claims 1 to 4, index detection is performed on each leaf node in the mobile communication network;

For the target leaf node whose detection result is that the key KPI index is abnormal, it is determined whether the target leaf node is a fault critical node based on the historical status of the target leaf node.
The method for locating faults in a mobile communication network according to claim 5, comprising: determining whether the target leaf node is a fault critical node in combination with the historical state of the target leaf node comprising:

When the historical state of the target leaf node is that the key KPI indicator is normal, determine that the target leaf node is not a fault critical node;

When the historical state of the target leaf node is that the key KPI index is abnormal, it is determined that the target leaf node is a fault critical node.
The method for locating faults in a mobile communication network according to claim 5 or 6, wherein, using the index detection model, performing index detection on each leaf node in the mobile communication network comprises:

The time sequence sequence of each leaf node is predicted by the index detection model, and when the difference between the predicted value and the actual value is greater than the set threshold, it is determined that the key KPI index of the target leaf node corresponding to the index detection model is abnormal.
The method for locating faults in a mobile communication network according to any one of claims 5 to 7, wherein, when there are multiple target leaf nodes determined to be critical fault nodes, the method further comprises:

Obtain the fluctuation contribution degree of each target leaf node determined as the fault key node, and the fluctuation contribution degree is a value that reflects the degree of influence on other leaf nodes connected to it when the target leaf node fails;

It is determined that the target leaf node with the largest fluctuation contribution is the fault root cause node.
The method for locating faults in a mobile communication network according to any one of claims 5 to 8, wherein, using the indicator detection model, before performing indicator detection on each leaf node in the mobile communication network, the method further comprises:

According to the obtained user plane signaling and control plane signaling, self-learning is performed on the topology of the mobile communication network, so as to obtain the topology of the mobile communication network, and each network in the obtained topology is Entities act as leaf nodes.
A device for obtaining an index detection model for a mobile communication network, comprising:

an acquisition module, configured to acquire key KPI indicators of each leaf node in the mobile communication network according to a preset key KPI indicator system, where the key KPI indicator system includes at least two types of key KPI indicators;

The training module is used for aggregating the key KPI indicators of each leaf node by category, extracting the characteristic parameters of each type of key KPI indicators aggregated by each leaf node, and dividing the same type of key KPI indicators of each leaf node. The characteristic parameters of the indicators are used as the input of the preset model for model training, and the indicator detection model corresponding to each type of key KPI indicators is obtained.
A fault location device for a mobile communication network, comprising:

An anomaly detection module, configured to use the index detection model obtained by the index detection model acquisition method for a mobile communication network according to any one of claims 1 to 4, to detect each leaf node in the mobile communication network perform index detection;

The fault locating module is used to determine whether the target leaf node is a fault critical node in combination with the historical status of the target leaf node for the target leaf node whose detection result is that the key KPI index is abnormal.
An electronic device comprising a processor, a memory, and a communication bus connecting the processor and the memory;

The memory stores a first computer program, and the first computer program can be executed by the processor to implement the method for obtaining an index detection model for a mobile communication network according to any one of claims 1 to 4 A step of;

and / or,

The memory stores a second computer program executable by the processor to implement the steps of the method for locating a fault in a mobile communication network according to any one of claims 5 to 9.
A computer-readable storage medium, the computer-readable storage medium stores a first computer program, the first computer program can be executed by a processor, to realize the use according to any one of claims 1 to 4. The steps of a method for obtaining an index detection model for a mobile communication network;

and / or,

The computer-readable storage medium stores a second computer program, and the second computer program can be executed by the processor to implement the steps of the method for locating a fault in a mobile communication network according to any one of claims 5 to 9 .