CN114765575B

CN114765575B - Network fault cause prediction method and device and electronic equipment

Info

Publication number: CN114765575B
Application number: CN202110001432.4A
Authority: CN
Inventors: 周永庆; 花小磊; 朱琳
Original assignee: China Mobile Communications Group Co Ltd; China Mobile Communications Ltd Research Institute
Current assignee: China Mobile Communications Group Co Ltd; China Mobile Communications Ltd Research Institute
Priority date: 2021-01-04
Filing date: 2021-01-04
Publication date: 2024-06-11
Anticipated expiration: 2041-01-04
Also published as: CN114765575A

Abstract

The invention provides a network failure cause prediction method, a network failure cause prediction device and electronic equipment, and solves the problem that the existing network failure cause prediction accuracy is low. The method of the invention comprises the following steps: obtaining a classification feature vector in a fault work order, wherein the classification feature vector comprises a first type feature vector and a second type feature vector; obtaining a target fault reason category to which the fault work order belongs according to the first class feature vector and the first classification prediction model; and obtaining the target fault source factor category in the target fault cause category by the fault worker Shan Zaisuo according to the second class feature vector and the second class prediction model corresponding to the target fault cause category. According to the method, the two-step prediction method is used for predicting the major fault cause class firstly and then predicting the class of the fault cause subdivision in the major fault cause class, so that the number of the classes predicted in each step can be effectively reduced, and the accuracy of the prediction result is improved.

Description

Network fault cause prediction method and device and electronic equipment

Technical Field

The present invention relates to the field of artificial intelligence technologies, and in particular, to a network failure cause prediction method and apparatus, and an electronic device.

Background

In a network system, network elements are various, a network structure is complex, and various faults are inevitably generated in the process of network operation. After the fault occurs, network operation staff needs to search the fault to find out the reason for the fault, and further, corresponding treatment measures are adopted to help the existing network to resume operation.

Specifically, in the running process of the existing network, after the fault occurs, the network equipment generates an alarm and reports the alarm to the network management system. The network management system sends a sheet to operation and maintenance personnel based on the received alarm and a certain sheet sending rule, the operation and maintenance personnel checks the fault reason by combining information in various aspects such as the alarm, and then adopts corresponding treatment measures according to the fault reason, and after the fault is solved, the fault reason and the treatment measure sheet are corresponding to the corresponding work sheet.

In the existing fault cause prediction technical scheme, the types of fault causes are more, some fault causes are similar, and the accuracy rate is lower when the prediction is directly carried out.

Disclosure of Invention

The invention aims to provide a network failure cause prediction method, a network failure cause prediction device and electronic equipment, which are used for solving the problem that the existing network failure cause prediction accuracy is low.

In order to achieve the above object, the present invention provides a network failure cause prediction method, including:

obtaining a classification feature vector in a fault work order, wherein the classification feature vector comprises a first type feature vector and a second type feature vector;

Obtaining a target fault reason category to which the fault work order belongs according to the first class feature vector and the first classification prediction model;

and obtaining a target fault source factor category in the target fault cause categories of the fault worker Shan Zaisuo according to the second class feature vector and a second class prediction model corresponding to the target fault category.

The obtaining the classification feature vector in the fault work order comprises the following steps:

acquiring a fault work order to be processed, wherein the fields of the fault work order comprise an alarm title, a network element name, a network element type and fault occurrence time;

And extracting the classified feature vectors in the fault worksheet based on the corresponding relation between the fields of the fault worksheet and the feature vectors and/or the feature extraction model.

Wherein the classification feature vector comprises:

a first feature vector for characterizing the alert title;

The second feature vector is used for representing the fault reason category corresponding to the alarm title;

A third feature vector for characterizing the network element type;

a fourth feature vector for characterizing a failure cause category corresponding to the network element type;

a fifth feature vector for characterizing alert information associated with the failed worksheet;

A sixth feature vector for characterizing a fault source factor class corresponding to the alarm header; a seventh feature vector for characterizing a failure origin factor class corresponding to the network element type;

wherein the first type of feature vector comprises: the first feature vector, the second feature vector, the third feature vector, the fourth feature vector, and the fifth feature vector;

The second class of feature vectors includes: the first feature vector, the second feature vector, the third feature vector, the fourth feature vector, the fifth feature vector, the sixth feature vector, and the seventh feature vector.

The obtaining, according to the first class feature vector and the first classification prediction model, the target fault cause class to which the fault work order belongs includes:

classifying the first class feature vectors through the first classification prediction model to obtain probability values of various fault cause classes;

And determining the fault reason category corresponding to the maximum probability value in the probability values of the fault reason categories as the target fault reason category to which the fault work order belongs.

The obtaining, according to the second class feature vector and a second class prediction model corresponding to the target fault cause class, a target fault source factor class in the target fault cause classes of the fault worker Shan Zaisuo includes:

Classifying the second class feature vector through the second classification prediction model to obtain probability values of each fault source factor class in the target fault cause class of the fault worker Shan Zaisuo;

and determining the fault original factor category corresponding to the maximum probability value in the probability values of the fault original factor categories as the target fault original factor category.

Wherein the method further comprises:

Acquiring a plurality of historical fault worksheets and a plurality of historical alarm messages, wherein the fields of each historical fault worksheet comprise an alarm title, a network element name, a network element type, fault occurrence time, a fault reason category and a fault source factor category corresponding to the fault reason category, and each field of the historical alarm messages comprises an alarm title, a network element name and an alarm starting time;

Obtaining a classification feature vector according to the field of the historical fault work order and the field of the historical alarm information, wherein the classification feature vector comprises: the method comprises the steps of representing a first feature vector of the alarm header, representing a second feature vector of a fault cause category corresponding to the alarm header, representing a third feature vector of the network element type, representing a fourth feature vector of the fault cause category corresponding to the network element type, representing a fifth feature vector of alarm information associated to the fault work order, representing a sixth feature vector of a fault source factor category corresponding to the alarm header and representing a seventh feature vector of the fault source factor category corresponding to the network element type;

And performing model training according to the first feature vector, the second feature vector, the third feature vector, the fourth feature vector, the fifth feature vector and the class labels of fault cause classes to obtain a first classification prediction model.

Wherein, after obtaining the classification feature vector according to the field of the historical fault work order and the field of the historical alarm information, the method further comprises:

Grouping a plurality of historical fault worksheets according to the fault cause categories to obtain a plurality of groups of historical fault worksheets;

and respectively carrying out model training on each group of historical fault work order data according to the class labels and the classification feature vectors of the fault source factor classes corresponding to each group of historical fault work order data to obtain a plurality of second classification prediction models.

The invention also provides a network failure cause prediction device, which comprises:

The first acquisition module is used for acquiring classification feature vectors in the fault worksheet, wherein the classification feature vectors comprise first-class feature vectors and second-class feature vectors;

The first fault reason prediction module is used for obtaining a target fault reason category to which the fault work order belongs according to the first class feature vector and the first classification prediction model;

And the second failure cause prediction module is used for obtaining a target failure source factor category in the target failure cause categories of the failure worker Shan Zaisuo according to the second class feature vector and a second class prediction model corresponding to the target failure cause category.

The present invention also provides an electronic device comprising a processor and a transceiver, the transceiver receiving and transmitting data under the control of the processor, the processor being configured to:

And obtaining the target fault source factor category in the target fault cause category of the fault worker Shan Zaisuo according to the second class feature vector and the second class prediction model.

Wherein the processor is further configured to:

Wherein the classification feature vector comprises:

a first feature vector for characterizing the alert title;

A third feature vector for characterizing the network element type;

A sixth feature vector for characterizing a fault source factor class corresponding to the alarm header; and

A seventh feature vector for characterizing a failure primitive class corresponding to the network element type;

Wherein the processor is further configured to:

The invention also provides an electronic device comprising a memory, a processor and a program stored on the memory and capable of running on the processor; the processor implements the network failure cause prediction method described above when executing the program.

The present invention also provides a computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps in a network failure cause prediction method as described above.

The technical scheme of the invention has at least the following beneficial effects:

In the embodiment of the invention, the classification feature vector in the fault work order is obtained, wherein the classification feature vector comprises a first type feature vector and a second type feature vector; obtaining a target fault reason category to which the fault work order belongs according to the first class feature vector and the first classification prediction model; according to the second class feature vector and the second class prediction model corresponding to the target fault cause class, the target fault source factor class in the target fault cause class of the fault worker Shan Zaisuo is obtained, so that the fault cause major class is predicted firstly by a two-step prediction method, and then the class of fault cause subdivision in the fault cause major class is predicted, the number of the classes predicted in each step can be effectively reduced, and the accuracy of a prediction result is improved.

Drawings

Fig. 1 shows one of flow diagrams of a network failure cause prediction method according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of a model training process of a first classification prediction model and a second classification prediction model according to an embodiment of the invention;

FIG. 3 is a second flow chart of a network failure cause prediction method according to an embodiment of the invention;

fig. 4 is a schematic block diagram of a network failure cause prediction apparatus according to an embodiment of the present invention;

fig. 5 shows a schematic structural diagram of an electronic device according to an embodiment of the present invention.

Detailed Description

In order to make the technical problems, technical solutions and advantages to be solved more apparent, the following detailed description will be given with reference to the accompanying drawings and specific embodiments.

Aiming at the problem of low accuracy of predicting the network failure cause, the invention provides a network failure cause predicting method, a network failure cause predicting device and electronic equipment.

Fig. 1 is a schematic flow chart of a network failure cause prediction method according to an embodiment of the present invention. The method specifically comprises the following steps:

Step 101, obtaining a classification feature vector in a fault work order, wherein the classification feature vector comprises a first type feature vector and a second type feature vector;

In the step, the fault work order is generated according to the preset dispatch rule based on the received alarm information, and is a new work order, namely the fault reason is not recorded on the fault work order temporarily.

Step 102, obtaining a target fault reason category to which the fault work order belongs according to the first class feature vector and the first classification prediction model;

in the step, the first classification prediction model is a pre-trained model, the first class feature vector of the fault work order is taken as input, the first classification prediction model is input, and the target fault cause category of the fault work order, namely the fault cause major category of the fault work order is output.

The first classification prediction model is trained based on historical fault worksheets and alarm data.

And step 103, obtaining a target fault source factor category in the target fault cause categories by the fault worker Shan Zaisuo according to the second class feature vector and a second class prediction model corresponding to the target fault cause category.

In this step, the second classification prediction model is a model trained in advance.

Here, the second class feature vector of the failure work order is taken as an input, the second class prediction model is input, and the target failure factor class of the failure work order in the target failure cause class, namely the class of failure cause subdivision in the failure cause major class, is output.

It should be noted that, because the historical fault worksheet includes the fault reasons and the processing measures recorded by the operation and maintenance personnel after the fault is resolved, and contains a lot of operation and maintenance experience, the machine learning model is trained according to the information contained in the historical fault worksheet to obtain the first classification prediction model and the second classification prediction model, and then the worksheet information and the alarm information in the new dispatching sheet automatically predict the fault reasons, so that the time for the operation and maintenance personnel to troubleshoot the fault reasons is saved.

According to the network fault cause prediction method, the classification feature vectors in the fault worksheet are obtained, wherein the classification feature vectors comprise a first type feature vector and a second type feature vector; obtaining a target fault reason category to which the fault work order belongs according to the first class feature vector and the first classification prediction model; according to the second class feature vector and the second class prediction model corresponding to the target fault cause class, the target fault source factor class in the target fault cause class of the fault worker Shan Zaisuo is obtained, so that the fault cause major class is predicted firstly by a two-step prediction method, and then the class of fault cause subdivision in the fault cause major class is predicted, the number of the classes predicted in each step can be effectively reduced, and the accuracy of a prediction result is improved.

As an alternative implementation manner, step 101 of the embodiment of the present invention may specifically include:

In this step, after the network failure occurs, an alarm message is generated. And the electronic equipment generates a fault work order to be processed according to a preset order form rule based on the received alarm information.

Typically, the failure worksheet to be processed includes, but is not limited to, fields such as an alarm header, a network element name, a network element type, and a failure occurrence time. It should be noted that, at this time, the fault reason causing the network fault and the corresponding processing measures to be taken are not recorded in the fault work order to be processed.

Optionally, the classification feature vector includes:

a first feature vector for characterizing the alert title;

A third feature vector for characterizing the network element type;

The first feature vector, the second feature vector and the sixth feature vector are obtained according to the corresponding relation between the alarm title and the feature vector of the fault work order.

That is, the alarm title and the characteristic vector representing the alarm title have a first corresponding relation, and the first characteristic vector representing the alarm title of the fault work order is obtained through the first corresponding relation; the alarm title and the feature vector representing the fault cause category corresponding to the alarm title have a second corresponding relation, and the second feature vector representing the fault cause category corresponding to the alarm title of the fault work order is obtained through the second corresponding relation; the alarm title and the feature vector representing the fault original factor category corresponding to the alarm title have a third corresponding relation, and a sixth feature vector representing the fault original factor category corresponding to the alarm title of the fault work order is obtained through the third corresponding relation.

The third feature vector, the fourth feature vector and the seventh feature vector are obtained according to the corresponding relation between the network element type of the fault work order and the feature vector.

That is, the network element type and the characteristic vector representing the network element type have a fourth corresponding relation, and a third characteristic vector representing the network element type of the fault work order is obtained through the fourth corresponding relation; the network element type and the feature vector representing the fault cause category corresponding to the network element type have a fifth corresponding relation, and a fourth feature vector representing the fault cause category corresponding to the network element type of the fault work order is obtained through the fifth corresponding relation; the network element type and the feature vector representing the fault original factor class corresponding to the network element type have a sixth corresponding relation, and a seventh feature vector representing the fault original factor class corresponding to the network element type of the fault work order is obtained through the sixth corresponding relation.

It should be noted that the fifth feature vector is obtained based on a feature extraction model. Specifically, firstly, extracting alarm information associated with the fault work order to obtain m alarms, and obtaining corresponding vectors for each alarm information through a first feature extraction model (such as CBOW models of word2 vec); then, a first average value vec_ cbow of m vectors is obtained; then, each piece of alarm information is subjected to a second feature extraction model (such as a Skip-gram model of word2 vec) to obtain a corresponding vector; then, obtaining a second average value vec_sg of m vectors; and finally, splicing the first average value vec_ cbow and the second average value vec_sg to obtain a fifth feature vector.

It should be noted that, when the alarm information satisfies the first condition and the second condition, the alarm information is determined to be the alarm information associated with the fault work order.

Here, the first condition is that the alarm start time t ₂ of the alarm information is between (t ₁-t_A,t₁+t_B), where t ₁ represents the failure occurrence time of the failure work order, and t _A and t _B are preset time values.

That is, the alarm start time of the alarm information is between the previous period and the subsequent period of the failure occurrence time of the failure work order.

The second condition is that the network element name in the alarm information is the same as the network element name in the fault work order.

Here, word2vec can be according to given corpus, through training model after optimizing, express a word into the vector form effectively fast, offer the new instrument for application study in the field of natural language processing. word2vec relies on skipping certain symbol Skip-grams models or continuous word bag CBOW models to build the neuropord embedding.

As an optional implementation manner, the method step 102 of the embodiment of the present invention obtains, according to the first class feature vector and the first classification prediction model, a target failure cause class to which the failure work order belongs, which may include:

here, each failure cause category, that is, each failure cause category to which the failure work order belongs.

It should be noted that, based on the first class feature vector, the fault cause major class to which the fault work order belongs may correspond to a plurality of classes through classification of the first classification prediction model, and the fault cause major class to which the fault work order most likely belongs may be determined through probability value comparison and measurement.

By the implementation mode, the target fault reason category to which the fault work order belongs can be predicted.

As an optional implementation manner, the method step 103 of the embodiment of the present invention obtains, according to the second class feature vector and a second class prediction model corresponding to the target fault cause class, a target fault source factor class in the target fault cause class of the fault worker Shan Zaisuo, where the method includes:

In this step, it should be noted that the second classification prediction model is determined according to the target failure cause classification to which the failure work order belongs. That is, different failure cause categories correspond to different second classification prediction models.

The implementation is similar to the above-mentioned determination of the target fault cause category of the fault worksheet, and the most likely fault source factor category can be determined by probability value comparison and measurement.

As can be seen from the above description, the accuracy of predicting the failure cause of the failure work order is critical to the classification prediction model, how to train the classification prediction model, and as an optional implementation manner, the method of the embodiment of the present invention may further include:

Here, "multiple" of the plurality of historical fault worksheets and the plurality of historical alert information may be understood herein as a number. It should be noted that, these historical fault worksheets and historical alarm information are valid data, that is, there is no empty field in the historical fault worksheets and the historical alarm information.

In a large number of historical fault worksheets and a large number of historical alarm information, firstly, removing data containing empty fields; then, regularized matching is carried out on the screened historical fault worksheets and alarm titles in the alarm information, alarm content is extracted, and a yyy alarm part is extracted if a yyy alarm occurs to 'xxx'; finally, the fault occurrence time and the alarm start time are formatted and converted into a unified time format, such as datetime's 64 format.

In the step, a first feature vector, a second feature vector and a sixth feature vector are obtained according to the alarm header field of the historical fault worksheet.

Specifically, the history fault worksheet is processed as follows:

1) Carrying out one-hot coding on the alarm title of the historical fault work order to obtain one-hot vectors; the alert title and the corresponding one-hot vector are stored as keys and values of a dictionary, which is denoted as subject_1.

The one-hot vector is the first eigenvector used to characterize the alarm header.

2) Obtaining the times of each major class of fault reasons which correspondingly occur on the alarm title in the historical fault work order, and carrying out normalization processing to obtain a vector corresponding to the alarm title; the alarm title and the corresponding vector are stored as keys and values of a dictionary, which is denoted as the dictionary_2.

Here, the vector is a second feature vector for characterizing the failure cause category corresponding to the alarm header.

For example, in all worksheets, an alarm header a appears in a total of 4 worksheets, and the alarm headers and the failure causes in the four worksheets are mainly as follows:

alarm header A failure cause class 1

Alarm header A failure cause class 1

Alarm header A failure cause category 3

Alarm header A failure cause category 4

The vector corresponding to the alarm header A is [2/4,0,1/4,1/4,0, ], and the dimension of the vector is the category number of the fault reason category.

3) Performing one-hot coding on the network element type of the historical fault work order to obtain one-hot vectors; the network element type and the corresponding one-hot vector are stored as keys and values of a dictionary, and the dictionary is named as a direct_3.

The one-hot vector is a third feature vector for representing the alarm header.

4) Obtaining the times of each major fault cause class correspondingly appearing on the network element types in the historical fault worksheet, and carrying out normalization processing to obtain vectors corresponding to the network element types; the network element type and the corresponding vector are stored as keys and values of a dictionary, and the dictionary is named as a direct_4.

Here, the vector is a fourth feature vector for characterizing a failure cause category corresponding to the network element type.

For example, in all worksheets, a total of 4 worksheets have network element type a, and the network element types and failure causes in the four worksheets are mainly as follows:

network element type A failure cause major class 1

Network element type A failure cause major class 1

Network element type a failure cause major class 3

Network element type a failure cause major class 4

The vector corresponding to the network element type A is [2/4,0,1/4,1/4,0, ], and the dimension of the vector is the category number of the failure cause categories.

5) For a fifth feature vector for characterizing alert information to which a faulty work order is associated

The failed worksheet herein refers to a historical failed worksheet.

First, for each historical trouble ticket, the alert information to which it is associated is extracted.

Here, when the alarm information satisfies the first condition and the second condition, the alarm information is determined to be the alarm information to which the history trouble work order is associated.

Here, the first condition is that the alarm start time t ₂ of the alarm information is between (t ₁-t_A,t₁+t_B), where t ₁ represents the failure occurrence time of the historical failure work order, and t _A and t _B are preset time values.

That is, the alarm start time of the alarm information is between the previous period and the subsequent period of the failure occurrence time of the history failure work order.

The second condition is that the network element name in the alarm information is the same as the network element name in the historical fault work order.

And then, sequencing the alarm information associated with each historical fault work order according to the order of the alarm starting time, extracting the alarm title of each alarm information, and forming an alarm sentence by taking the alarm title as a word.

Here, the alert statement is used to describe a series of ordered alert messages generated on the network element within a period of time when the historical failure work order failure occurs.

Then, forming an alarm statement corresponding to each historical fault work order into a document which is used as a corpus, and respectively training CBOW models and Skip-gram models into wrod vec models by using the corpus to obtain two vector representations of each piece of alarm information; saving the trained CBOW model model_ cbow and Skip-gram model model_sg;

And finally, inquiring the alarm vector of the alarm title in the alarm statement corresponding to each historical fault work order, and then averaging the alarm vectors. The CBOW model and Skip-gram model vector and average the alarm headers, respectively. And splicing the two vectors obtained by the two models to be used as a feature vector of the history fault work order matched with the alarm information, namely a fifth feature vector.

6) Obtaining the times of each fault factor category (namely the category of fault reason subdivision) which correspondingly appears on the alarm title in the historical fault work order, and carrying out normalization processing to obtain a vector corresponding to the alarm title; the alarm title and the corresponding vector are stored as key and value, and the dictionary is named as the direct_5.

Here, the vector is a sixth feature vector for characterizing the fault source factor class corresponding to the alarm header.

7) Obtaining the times of each fault original factor category (namely the category of fault reason subdivision) which correspondingly appears on the network element types in the historical fault worksheet, and carrying out normalization processing to obtain vectors corresponding to the network element types; the network element type and the corresponding vector are stored as keys and values, and the dictionary is named as a direct_6.

Here, the vector is a seventh feature vector for characterizing the failure origin factor category corresponding to the network element type.

Optionally, the class label of the fault reason class is obtained by digitally encoding the fault reason class of the historical fault worksheet. I.e. the numbers 1 to N are used to identify N types of fault cause categories as category labels.

Here, the first feature vector, the second feature vector, the third feature vector, the fourth feature vector and the fifth feature vector are taken as inputs, and are input into a preset classification model, a classification result, namely, a class label is output, the class label in the classification result is compared with a corresponding actual class label, parameters in the preset classification model are continuously adjusted, the difference between the class label in the classification result and the corresponding actual class label is reduced, the difference is reduced to a preset range, or the preset classification model reaches a minimum convergence position.

Here, the preset classification model used in model training is GBDT model or XGBoost model.

Here, GBDT (Gradient Boosting Decision Tree ) model is an additive model that trains a set of CART (Classification and Regression Trees ) serially, and finally sums the predictions of all regression trees, thus yielding a strong learner, each new tree fitting the negative gradient direction of the current loss function.

XGBoost (Extreme Gradient Boosting, gradient-lifted tree) model, again generating models serially, taking the sum of all models as output.

Here, the trained first classification prediction model model_1 is saved.

Further, after obtaining the classification feature vector according to the field of the historical fault work order and the field of the historical alarm information, the method further comprises:

here, all the data, i.e. all the historical fault worksheets, are grouped according to the fault cause categories, so as to obtain a plurality of groups of historical fault worksheet data.

Here, the historical fault worksheets of different groups correspond to different fault cause categories.

It should be noted that, assuming that the types of fault causes are N, a ¹,···,A^N, then the types of fault causes, that is, the types of fault causes are all a ¹, the labels of the first group of data are the types of fault source factors, and assuming that the types of fault source factors corresponding to the types of fault causes a ¹, that is, the types of fault cause subdivision are n_1, then the labels of the first group of data areTo/>N_1 tags in total.

Here, the classification feature vector specifically refers to a first feature vector, a second feature vector, a third feature vector, a fourth feature vector, a fifth feature vector, a sixth feature vector, and a seventh feature vector.

And training a model through each group of historical fault worksheets, namely classifying feature vectors in each group of historical fault worksheets respectively, wherein a preset classifying model used in model training is GBDT models or XGBoost models.

Here, the trained N second classification prediction models model_2_1 to model_2_N are saved.

Here, a specific training process of the first classification prediction model and the second classification prediction model may refer to fig. 2.

The implementation of the method according to the embodiment of the present invention will be specifically described with reference to an example, as shown in fig. 3.

S1: and receiving a work order to be predicted.

Here, the work order to be predicted is a new work order, and the work order to be predicted includes four fields of an alarm title, a network element name, a network element type and a fault occurrence time.

It should be noted that, based on the four fields and the alarm information associated with the work order, the field of the failure origin factor type (i.e. failure cause subdivision type) of the work order to be predicted is predicted.

S2: and extracting the feature vector of the first class classification of the work order.

Specifically, 1) one-hot feature of alert title: and taking the alarm title in the work order as a key, and inquiring a feature vector vec_1 corresponding to the alarm title in the dictionary part_1, namely a first feature vector.

2) Fault cause large-class distribution characteristics of alarm titles: and taking the alarm title in the work order as a key, and inquiring a feature vector vec_2 corresponding to the alarm title in the dictionary part_2, namely a second feature vector.

3) One-hot feature of network element type: and taking the network element type of the work order as a key, and inquiring a feature vector vec_3 corresponding to the network element type in the dictionary part_3, namely a third feature vector.

4) Fault cause large-class distribution characteristics of network element types: and taking the network element type of the work order as a key, and inquiring a feature vector vec_4 corresponding to the network element type in the dictionary part_4, namely a fourth feature vector.

5) The worksheet is associated with the word2vec feature of the alert:

Firstly, extracting alarm information associated with the work order, wherein the extraction method is the same as that in the training stage. Assuming that h alarms are associated, for each alarm, using wrod < 2 > vec model CBOW model to obtain vectors of the alarms, and then solving the average vec_ cbow of the h vectors; then, for each alarm, using a Skip-gram model to obtain a vector of the alarm, and solving a mean value vec_sg of h vectors; finally, the average value vec_ cbow and the average value vec_sg are spliced to obtain a feature vector vec_5, namely a fifth feature vector.

6) Distribution characteristics of fault cause subdivision on alarm headers: and taking the alarm title in the work order as a key, and inquiring the feature vector vec_6 corresponding to the alarm title in the dictionary part_5.

7) Distribution characteristics of fault cause subdivision on network element type: and inquiring the feature vector vec_7 corresponding to the network element type in the dictionary part_6 by taking the network element type in the work order as a key.

Here, the above-described vec_1, vec_2, vec_3, vec_4, and vec_5 belong to one-level classification of feature vectors, i.e., the first-type feature vectors in the above-described embodiment.

S3: the feature vector of the first class classification is input to the model_1 for classification.

Here, the extracted feature vectors vec_1, vec_2, vec_3, vec_4, vec_5 are spliced, and then input to the model_1 for classification, so as to obtain probabilities p_1, p·, p_n belonging to the major classes of the failure causes.

S4: and selecting a fault cause major class i with highest probability in the prediction result.

The prediction result is the probability p_1, p_N of the major class of each failure cause, and the class i (i is more than or equal to 1 and less than or equal to N) with the maximum probability value is the class to which the work order belongs in the first step of classification.

S5: and extracting the feature vector of the secondary classification of the work order.

The above-described vec_1, vec_2, vec_3, vec_4, vec_5, vec_6, and vec_7 belong to the two-level classification of feature vectors, i.e., the second-level feature vector in the above-described embodiment. The specific extraction process is described in the section S2, and will not be described here again.

S6: the feature vector of the secondary classification is input to model_2_i for classification.

Here, the extracted feature vectors vec_1, vec_2, vec_3, vec_4, vec_5, vec_6, and vec_7 are spliced, and then input to a model_2_i for classification, so as to obtain probabilities of subdividing each failure cause in the failure cause major class i

S7: and selecting fault reason subdivision j with highest probability in the prediction result.

Wherein the prediction result is the probability of each fault reason subdivision in the fault reason major class i The class j (j is more than or equal to 1 and less than or equal to n_i) with the largest probability value is the class to which the work order belongs in the second step of classification, namely the final fault reason subdivision class.

As shown in fig. 4, the embodiment of the present invention further provides a network failure cause prediction apparatus, where the apparatus includes:

A first obtaining module 401, configured to obtain a classification feature vector in a fault work order, where the classification feature vector includes a first class feature vector and a second class feature vector;

A first failure cause prediction module 402, configured to obtain, according to the first class feature vector and the first classification prediction model, a target failure cause class to which the failure work order belongs;

And a second failure cause prediction module 403, configured to obtain a target failure factor category in the target failure cause categories of the failure worker Shan Zaisuo according to the second class feature vector and a second class prediction model corresponding to the target failure cause category.

Optionally, the first obtaining module 401 includes:

The first acquisition unit is used for acquiring a fault work order to be processed, wherein the fields of the fault work order comprise an alarm title, a network element name, a network element type and fault occurrence time;

and the feature extraction unit is used for extracting the classified feature vectors in the fault worksheet based on the corresponding relation between the fields of the fault worksheet and the feature vectors and/or the feature extraction model.

Optionally, the classification feature vector includes:

a first feature vector for characterizing the alert title;

A third feature vector for characterizing the network element type;

Optionally, the first failure cause prediction module 402 includes:

The first processing unit is used for classifying the first class feature vectors through the first classification prediction model to obtain probability values of various fault cause classes;

And the second processing unit is used for determining the fault reason category corresponding to the maximum probability value in the probability values of the fault reason categories as the target fault reason category to which the fault work order belongs.

Optionally, the second failure cause prediction module 403 includes:

The third processing unit is configured to classify the second class feature vector by using the second classification prediction model, so as to obtain probability values of each fault source factor class in the target fault cause class of the fault worker Shan Zaisuo;

And the fourth processing unit is used for determining the fault original factor category corresponding to the maximum probability value in the probability values of the fault original factor categories as the target fault original factor category.

Optionally, the apparatus further comprises:

The second acquisition module is used for acquiring a plurality of historical fault worksheets and a plurality of historical alarm information, wherein the fields of each historical fault worksheet comprise an alarm title, a network element name, a network element type, a fault occurrence time, a fault reason category and a fault source factor category corresponding to the fault reason category, and each field of the historical alarm information comprises an alarm title, a network element name and an alarm starting time;

The first processing module is configured to obtain a classification feature vector according to the field of the historical fault work order and the field of the historical alarm information, where the classification feature vector includes: the method comprises the steps of representing a first feature vector of the alarm header, representing a second feature vector of a fault cause category corresponding to the alarm header, representing a third feature vector of the network element type, representing a fourth feature vector of the fault cause category corresponding to the network element type, representing a fifth feature vector of alarm information associated to the fault work order, representing a sixth feature vector of a fault source factor category corresponding to the alarm header and representing a seventh feature vector of the fault source factor category corresponding to the network element type;

And the first model training module is used for carrying out model training according to the first feature vector, the second feature vector, the third feature vector, the fourth feature vector, the fifth feature vector and the class labels of fault cause classes to obtain a first classification prediction model.

Optionally, the apparatus further comprises:

The second processing module is used for grouping a plurality of historical fault worksheets according to the fault cause categories to obtain a plurality of groups of historical fault worksheets;

The second model training module is used for respectively carrying out model training on each group of historical fault work order data according to the class labels and the classification feature vectors of the fault source factor classes corresponding to each group of historical fault work order data to obtain a plurality of second classification prediction models.

According to the network fault cause prediction device, the classification feature vectors in the fault worksheet are obtained, wherein the classification feature vectors comprise a first type feature vector and a second type feature vector; obtaining a target fault reason category to which the fault work order belongs according to the first class feature vector and the first classification prediction model; according to the second class feature vector and the second class prediction model corresponding to the target fault cause class, the target fault source factor class in the target fault cause class of the fault worker Shan Zaisuo is obtained, so that the fault cause major class is predicted firstly by a two-step prediction method, and then the class of fault cause subdivision in the fault cause major class is predicted, the number of the classes predicted in each step can be effectively reduced, and the accuracy of a prediction result is improved.

It should be noted that, the above device provided in the embodiment of the present invention can implement all the method steps implemented in the method embodiment and achieve the same technical effects, and detailed descriptions of the same parts and beneficial effects as those in the method embodiment in this embodiment are omitted.

In order to better achieve the above objects, as shown in fig. 5, an embodiment of the present invention further provides an electronic device, including a processor 500 and a transceiver 510, where the transceiver 510 receives and transmits data under the control of the processor, and the processor 500 is configured to perform the following procedures:

And obtaining a target fault source factor category in the target fault cause categories by the fault worker Shan Zaisuo according to the second class feature vector and a second class prediction model corresponding to the target fault cause category.

Optionally, the processor 500 is further configured to:

Optionally, the classification feature vector includes:

a first feature vector for characterizing the alert title;

A third feature vector for characterizing the network element type;

Optionally, the processor 500 is further configured to:

According to the electronic equipment provided by the embodiment of the invention, the classification feature vectors in the fault worksheet are obtained, and the classification feature vectors comprise the first type feature vectors and the second type feature vectors; obtaining a target fault reason category to which the fault work order belongs according to the first class feature vector and the first classification prediction model; and obtaining the target fault source factor category in the target fault cause categories of the fault worker Shan Zaisuo according to the second class feature vector and the second class prediction model, so that the two-step prediction method is used for predicting the major fault cause category and predicting the subdivided category of the fault cause in the major fault cause category, the number of the categories predicted in each step can be effectively reduced, and the accuracy of the prediction result is improved.

The embodiment of the invention also provides an electronic device, which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor realizes each process in the network fault cause prediction method embodiment as described above when executing the program, and can achieve the same technical effect, and in order to avoid repetition, the description is omitted.

The embodiment of the present invention also provides a computer readable storage medium, on which a computer program is stored, where the program when executed by a processor implements each process in the network failure cause prediction method embodiment described above, and the same technical effects can be achieved, and for avoiding repetition, a detailed description is omitted herein. The computer readable storage medium is, for example, a Read-Only Memory (ROM), a random access Memory (Random Access Memory RAM), a magnetic disk or an optical disk.

It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-readable storage media (including, but not limited to, magnetic disk storage and optical storage, etc.) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may 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, create means for implementing the functions specified in the flowchart flow or flows and/or block or blocks.

These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing apparatus 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 implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

While the foregoing is directed to the preferred embodiments of the present invention, it will be appreciated by those skilled in the art that various modifications and adaptations can be made without departing from the principles of the present invention, and such modifications and adaptations are intended to be comprehended within the scope of the present invention.

Claims

1. A network failure cause prediction method, comprising:

Obtaining a target fault source factor category in the target fault cause categories of the fault worker Shan Zaisuo according to the second class feature vector and a second class prediction model corresponding to the target fault cause category;

The method further comprises the steps of:

2. The method of claim 1, wherein the obtaining the classification feature vector in the failed worksheet comprises:

3. The method of claim 2, wherein the classification feature vector comprises:

a first feature vector for characterizing the alert title;

A third feature vector for characterizing the network element type;

4. The method of claim 1, wherein the obtaining, from the first class feature vector and the first classification prediction model, the target failure cause class to which the failure work order belongs comprises:

5. The method of claim 1, wherein the obtaining a target fault factor class of the target fault cause classes for the faulty worker Shan Zaisuo based on the second class feature vector and a second class prediction model corresponding to the target fault cause class comprises:

6. The method of claim 1, wherein after deriving the classification feature vector from the field of the historical fault worksheet and the field of the historical alert information, the method further comprises:

7. A network failure cause prediction apparatus, comprising:

The second fault cause prediction module is configured to obtain a target fault source factor class in the target fault cause classes of the fault worker Shan Zaisuo according to the second class feature vector and a second class prediction model corresponding to the target fault cause class;

8. An electronic device comprising a processor and a transceiver, the transceiver receiving and transmitting data under control of the processor, the processor being configured to:

Obtaining a target fault source factor category in the target fault cause categories of the fault worker Shan Zaisuo according to the second class feature vector and the second class prediction model;

the processor is further configured to:

9. The electronic device of claim 8, wherein the processor is further configured to:

10. The electronic device of claim 9, wherein the classification feature vector comprises:

a first feature vector for characterizing the alert title;

A third feature vector for characterizing the network element type;

11. The electronic device of claim 8, wherein the processor is further configured to:

12. The electronic device of claim 8, wherein the processor is further configured to:

13. The electronic device of claim 8, wherein the processor is further configured to:

14. An electronic device comprising a memory, a processor, and a program stored on the memory and executable on the processor; the network failure cause prediction method according to any one of claims 1 to 6 is implemented when the processor executes the program.

15. A computer-readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the steps in the network failure cause prediction method according to any one of claims 1 to 6.