CN117376087A - Method, device, equipment and storage medium for delimiting network quality problems - Google Patents
Method, device, equipment and storage medium for delimiting network quality problems Download PDFInfo
- Publication number
- CN117376087A CN117376087A CN202311090496.1A CN202311090496A CN117376087A CN 117376087 A CN117376087 A CN 117376087A CN 202311090496 A CN202311090496 A CN 202311090496A CN 117376087 A CN117376087 A CN 117376087A
- Authority
- CN
- China
- Prior art keywords
- data
- positioning
- delimitation
- model
- network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/06—Management of faults, events, alarms or notifications
- H04L41/0631—Management of faults, events, alarms or notifications using root cause analysis; using analysis of correlation between notifications, alarms or events based on decision criteria, e.g. hierarchy, tree or time analysis
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/142—Network analysis or design using statistical or mathematical methods
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Signal Processing (AREA)
- Mathematical Physics (AREA)
- Evolutionary Biology (AREA)
- Computer Networks & Wireless Communication (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Mathematical Analysis (AREA)
- Algebra (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The application relates to the technical field of communication, and provides a method, a device, equipment and a storage medium for delimiting network quality problems, wherein the method comprises the following steps: acquiring positioning data and delimitation data based on network data of at least one data source; inputting the positioning data into a positioning model to obtain positioning result data output by the positioning model; if the network quality difference problem exists based on the positioning result data, the positioning result data and the delimitation data are input into the delimitation model, and quality difference reason data output by the delimitation model are obtained. According to the embodiment of the application, the intelligent recognition, positioning and delimitation of the network quality problem are realized by adopting the positioning model and the delimitation model, a large amount of network data can be analyzed more quickly and efficiently by the model, the problem is accurately positioned, the manual investigation time is reduced, and the accuracy and the efficiency of the positioning and delimitation of the network quality problem are improved.
Description
Technical Field
The present disclosure relates to the field of communications technologies, and in particular, to a method, an apparatus, a device, and a storage medium for delimiting a network quality problem.
Background
Currently, the internet is evolving from a popular and publicized service to a direction of a popular and personalized service. The traditional network configuration, operation and maintenance modes based on the static strategy mode are limited by the limitation of cognition on a complex network, are difficult to adapt to the complexity of the current network service in the aspects of flexibility, instantaneity and the like, and cannot meet the network quality problem positioning and delimitation requirements in a complex network environment, so that the accuracy of the network quality problem positioning and delimitation is low.
Disclosure of Invention
The utility model provides a network quality problem delimits method, device, equipment and storage medium for solve the problem that network quality problem location delimits the accuracy low, through adopting location model and delimit the intelligent recognition location delimit that the model realized network quality problem, can analyze a large amount of network data more fast, high-efficient through the model, and pinpoint the problem place, reduced manual investigation time, improve the accuracy and the efficiency of network quality problem location delimit.
The application provides a method for delimiting network quality problems, which comprises the following steps:
acquiring positioning data and delimitation data based on network data of at least one data source;
inputting the positioning data into a positioning model to obtain positioning result data output by the positioning model; the positioning model is obtained by training sample positioning data;
if the network quality difference problem exists based on the positioning result data, inputting the positioning result data and the delimitation data into a delimitation model, and obtaining quality difference reason data output by the delimitation model; the delimitation model is obtained by training sample positioning result data and sample delimitation data.
In one embodiment, the positioning model is trained based on the following steps:
collecting the sample positioning data;
performing data annotation on the sample positioning data to obtain training data;
preprocessing the training data, wherein the preprocessing comprises missing value processing and outlier processing;
and training the preprocessed training data by adopting a first target algorithm to obtain the positioning model.
In one embodiment, the performing data labeling on the sample positioning data to obtain training data includes:
obtaining a quality difference sample and a non-quality difference sample of the sample positioning data;
based on a set labeling strategy or a labeling model, carrying out data labeling on the quality difference sample;
and determining the training data based on the marked quality difference sample and the marked non-quality difference sample.
In one embodiment, the bounding model is trained based on the following steps:
collecting the sample positioning result data and the sample delimitation data;
and training the sample positioning result data and the sample delimitation data by adopting a second target algorithm to obtain the delimitation model.
In one embodiment, before the acquiring the positioning data and the delimiting data based on the network data of the at least one data source, the method further comprises:
constructing a warehouse entry table;
configuring a data processing task based on a data docking mode, information of a data source to which data belongs, task execution information and a task flow chart;
and accessing network data of at least one data source to a data management system based on the warehouse entry table and the data processing task.
In one embodiment, after the network data of at least one data source is accessed to the data management system based on the warehouse entry table and the data processing task, the method further comprises:
extracting performance indexes of the positioning data, and associating the performance indexes with network objects to generate a data table with the performance indexes associated with the network objects;
root cause data of the delimited data is extracted, and the root cause data is associated with a solution library to generate delimited conclusion data and a solution.
In one embodiment, determining that a network quality problem exists based on the positioning result data comprises:
acquiring quality difference time, quality difference objects and quality difference indexes based on the positioning result data;
and determining that a network quality difference problem exists based on the quality difference time, the quality difference object and the quality difference index.
The application provides a network quality problem delimits device, includes:
the data acquisition module is used for acquiring positioning data and delimitation data based on network data of at least one data source;
the positioning module is used for inputting the positioning data into a positioning model and obtaining positioning result data output by the positioning model; the positioning model is obtained by training sample positioning data;
the delimiting module is used for inputting the positioning result data and the delimiting data into a delimiting model if the network quality difference problem exists based on the positioning result data, and obtaining quality difference cause data output by the delimiting model; the delimitation model is obtained by training sample positioning result data and sample delimitation data.
The application also provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the network quality problem delimitation method as described in any of the above when executing the program.
The present application also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a network quality problem delimitation method as described in any of the above.
The method, the device, the equipment and the storage medium for delimiting the network quality problem acquire positioning data and delimiting data through network data based on at least one data source; inputting the positioning data into a positioning model to obtain positioning result data output by the positioning model; if the network quality difference problem exists based on the positioning result data, the positioning result data and the delimitation data are input into the delimitation model, and quality difference reason data output by the delimitation model are obtained. According to the embodiment of the application, the intelligent recognition, positioning and delimitation of the network quality problem are realized by adopting the positioning model and the delimitation model, a large amount of network data can be analyzed more quickly and efficiently by the model, the problem is accurately positioned, the manual investigation time is reduced, and the accuracy and the efficiency of the positioning and delimitation of the network quality problem are improved.
Drawings
For a clearer description of the present application or of the prior art, the drawings that are used in the description of the embodiments or of the prior art will be briefly described, it being apparent that the drawings in the description below are some embodiments of the present application, and that other drawings may be obtained from these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic flow chart of a network quality problem delimiting method provided in the present application;
FIG. 2 is a schematic block diagram of a network quality problem delimitation method package provided by the present application;
FIG. 3 is a schematic diagram of a network quality problem delimiter apparatus provided in the present application;
fig. 4 is a schematic structural diagram of an electronic device provided in the present application.
Detailed Description
For the purposes of making the objects, technical solutions and advantages of the present application more apparent, the technical solutions in the present application will be clearly and completely described below with reference to the drawings in the present application, and it is apparent that the described embodiments are some, but not all, embodiments of the present application. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments herein without making any inventive effort, are intended to be within the scope of the present application.
The network quality problem delimitation method, apparatus, device and storage medium of the present application are described below in connection with fig. 1-4.
Specifically, the present application provides a method for delimiting a network quality problem, and referring to fig. 1, fig. 1 is a schematic flow chart of the method for delimiting a network quality problem provided in the present application.
The method for delimiting the network quality problem provided by the embodiment of the application comprises the following steps:
step 100, acquiring positioning data and delimitation data based on network data of at least one data source;
it should be noted that, network data of different data sources are accessed into the home system of the capacity of locating and delimiting from the nano tube to the global quality end-to-end, based on which, network data of different data sources, such as signaling data, OMC (Operation and Maintenance Center, operation and maintenance and management center) northbound data, alarm data, dial-up measurement data, log data, error code solution library, alarm solution library, can be obtained from the home system; and then, acquiring positioning data and delimitation data based on the network data, wherein the positioning data comprises signaling data, OMC northbound data and the like, and the delimitation data comprises alarm data, dial test data, log data, an error code solution library, an alarm solution library and the like.
Step 200, inputting the positioning data into a positioning model, and obtaining positioning result data output by the positioning model; the positioning model is obtained by training sample positioning data;
the positioning model is obtained by training sample positioning data, and positioning result data output by the positioning model can be obtained by inputting the positioning data into the positioning model, wherein the positioning result data comprises quality difference time, quality difference objects and quality difference indexes, and the positioning result data represents the time period of occurrence of problems, the affected objects and the specific quality difference indexes.
Step 300, if it is determined that the network quality difference problem exists based on the positioning result data, inputting the positioning result data and the delimiting data into a delimiting model, and obtaining quality difference cause data output by the delimiting model; the delimitation model is obtained by training sample positioning result data and sample delimitation data.
And determining that the network quality difference problem exists based on the positioning result data, specifically, acquiring quality difference time, quality difference objects and quality difference indexes based on the positioning result data, and then determining that the network quality difference problem exists based on the quality difference time, the quality difference objects and the quality difference indexes. For example, the bad time indicates a period of time in which network performance is abnormal or degraded, and by examining the performance data changes during the bad period of time, it can be determined whether there is a network bad problem. The bad object represents an affected network node, device, service area, or other related entity, and by determining the bad object, the problem area can be narrowed down and targeted for location and resolution. The quality difference index reflects specific aspects of poor network performance, and the indexes can comprise delay, packet loss rate, bandwidth utilization rate and the like, and by analyzing the change and trend of the quality difference index, specific performance problems causing poor network quality can be determined. By comprehensively considering the quality difference time, the quality difference object and the quality difference index, the time period of occurrence of the problem, the affected object and the specific index problem causing the network quality difference can be positioned. Based on the method, the network quality difference problem can be identified, positioned and solved quickly, so that the network service quality and the user experience can be improved.
If the network problem is determined to exist, the reason for the problem needs to be further determined, at this time, the positioning result data and the delimitation data need to be input into a delimitation model, and quality difference reason data output by the delimitation model is obtained, wherein the delimitation model is obtained by training sample positioning result data and sample delimitation data. For example, the AI algorithm is invoked by the incoming capability entry to achieve the bounding conclusion output. The capacity input parameter is divided into two parts, namely a delimitation object which is output data of the positioning model, namely quality difference time, quality difference object and quality difference index; secondly, error code data, alarm data, dial test data, log data and the like which assist in carrying out delimitation conclusion output are generated by associating the two types of input parameters into multi-dimensional potential cause data of a quality difference object, contribution ratio calculation is carried out by using an AI algorithm, and the first N quality difference causes are output to guide problem repair.
According to the network quality problem delimiting method, positioning data and delimiting data are obtained through network data based on at least one data source; inputting the positioning data into a positioning model to obtain positioning result data output by the positioning model; the positioning model is obtained by training sample positioning data; if the network quality difference problem exists based on the positioning result data, inputting the positioning result data and the delimitation data into a delimitation model, and obtaining quality difference reason data output by the delimitation model; the delimitation model is obtained by training sample positioning result data and sample delimitation data. According to the embodiment of the application, the intelligent recognition, positioning and delimitation of the network quality problem are realized by adopting the positioning model and the delimitation model, a large amount of network data can be analyzed more quickly and efficiently by the model, the problem is accurately positioned, the manual investigation time is reduced, and the accuracy and the efficiency of the positioning and delimitation of the network quality problem are improved.
Based on the above embodiment, the positioning model is trained based on the following steps:
step 210, collecting the sample positioning data;
220, performing data annotation on the sample positioning data to obtain training data;
step 230, preprocessing the training data, wherein the preprocessing comprises missing value processing and outlier processing;
and step 240, training the preprocessed training data by adopting a first target algorithm to obtain the positioning model.
Sample positioning data such as signaling data, OMC northbound data and the like are collected, and then the sample positioning data is processed to form wide table data integrating signaling performance and OMC northbound performance of the whole object.
And carrying out data labeling on the sample positioning data in the wide-table data to obtain training data. Specifically, a quality difference sample and a non-quality difference sample of sample positioning data are obtained; based on a set labeling strategy or a set labeling model, carrying out data labeling on the quality difference sample; and determining training data based on the marked quality difference sample and the marked non-quality difference sample. For example, the set labeling strategy can be a labeling strategy determined based on expert experience, that is, based on expert experience labeling, the labeling method is based on experience and knowledge of field experts, and judges whether the quality difference standard is reached by comparing the data of the performance index with a set threshold value, mainly based on the quality difference object labeling from the performance index level by comparing the key performance index threshold value and the index with the ring ratio fluctuation threshold value, and simultaneously, the alarm object, the dial test failure object and the log error object are all labeled as quality difference objects, and a primary global problem object library is constructed. The labeling model can be labeled by an AI algorithm, mainly based on multi-source data of performance, alarm, dial test and log, and the AI algorithm is introduced to learn, so that the quality of a labeling object is poor. Based on the quality difference sample marking is carried out on the constructed global problem object library through expert experience marking or AI algorithm marking, and meanwhile, non-quality difference samples with the same time period are selected to generate training data with quality difference samples and non-quality difference samples.
Performing missing value processing and outlier processing on the training data, wherein the missing value processing comprises the following steps:
deleting the feature containing the missing value: if the deletion rate of the variable is higher (more than 80%), the coverage rate is lower, and the importance is lower, the variable can be directly deleted.
And (5) mean value interpolation: the attribute of the data is divided into a fixed-distance type and a non-fixed-distance type, and if the missing value is fixed-distance type, the missing value is interpolated by the average value of the existence values of the attribute; if the missing values are non-fixed, the missing values are complemented with the mode of the attribute (i.e. the value with the highest frequency of occurrence) according to the mode principle in statistics.
Maximum likelihood estimation: under the condition that the deletion type is random deletion, assuming that the model is correct for a complete sample, the unknown parameters can be estimated with maximum likelihood by observing the marginal distribution of the data.
Interpolation method filling comprises random interpolation, multiple difference interpolation method, lagrange interpolation, newton interpolation and the like.
Outlier processing includes:
absolute value difference median method: the method of detecting outliers by summing the distances between all factors and the average value is first calculated. Logic for processing: step one, finding out the median Xmedia of all factors; step two, obtaining an absolute deviation value Xi-Xmedia of each factor and a median; thirdly, obtaining a median MAD of the absolute deviation value; finally, the parameter n is determined, so that a reasonable range is determined as [ Xmedian-nMAD, xmedian+nmad ], and an adjustment is made for the factor values out of the reasonable range.
3 sigma standard deviation method: the standard deviation itself may represent the degree of dispersion of the factors, based on the mean of the factors Xmean. During outlier processing, the distance of the factor from the mean can be measured by Xmean+ -nσ.
Percentile method: the factor values are sorted in ascending order, with factor values higher than 97.5% in the ranking percentile or lower than 2.5% in the ranking percentile.
Clustering: small clusters far from other clusters are discarded using a clustering algorithm.
After preprocessing the training data, algorithm selection and training tuning are performed. For example, analyzing data characteristics and capability requirements, constructing a quality end-to-end positioning model by adopting an autoregressive moving average algorithm, a discrete wavelet transform algorithm, a DWAFE algorithm and a time sequence prediction algorithm based on a real-propset, and selecting a final algorithm according to a model result, wherein the finally selected algorithm is a first target algorithm. Training the preprocessed training data by adopting a first target algorithm to obtain a positioning model. Meanwhile, parameter adjustment is reversely carried out based on each training effect, the algorithm effect is continuously optimized, the optimal parameters are determined, and the model effect is ensured.
According to the embodiment of the application, an automatic problem positioning process can be realized by training the positioning model. Compared with the traditional manual positioning method, the model can analyze a large amount of network data more quickly and efficiently and accurately position the problem, so that the time and effort for manual troubleshooting and trial-and-error are reduced, and the problem processing efficiency and the network service quality are improved.
Based on the above embodiment, the bounding model is trained based on the following steps:
step 310, collecting the sample positioning result data and the sample delimitation data;
and 320, training the sample positioning result data and the sample delimitation data by adopting a second target algorithm to obtain the delimitation model.
And acquiring a historical output result of the positioning model, taking the historical output result as a sample positioning result, simultaneously acquiring processed historical delimiting data from a data management system as sample delimiting data, and training the sample positioning result data and the sample delimiting data by adopting a second target algorithm to obtain a delimiting model.
For example, using a transducer as an example for the second target algorithm, the training of the bounding model is described in an analytical way:
data preparation: sample data required for the delimitation task, such as sample positioning result data and sample delimitation data, are collected and arranged. These sample data should contain the input sequence (such as text or other sequence data) and the corresponding delimiter tag. The delimiting label indicates a problem location or an error location in the sample data.
Data preprocessing: the sample data is preprocessed, including text segmentation, sequence filling and the like, so that all the sample data are ensured to have the same length, and model training is facilitated.
Constructing a transducer model: and constructing a network structure of the transducer model. The transducer model is composed of multiple encoder and decoder layers, each of which includes a self-attention mechanism and a feed-forward neural network. The method can be implemented by using an existing transducer model, such as BERT (Bidirectional Encoder Representations from Transformers, pre-training language model), GPT (generating Pre-trained Transformer, pre-training language model), and the like, or can be built by self.
Model training: the transducer model is trained using the prepared sample data. By using the input sequence as the input of the model, and the delimitation label as the target output, the model parameters are adjusted through an optimization method (such as back propagation and gradient descent), so that the model can learn the delimitation mode in the sample data.
Model evaluation and tuning: model performance during training is evaluated using a validation set or cross-validation method. And performing model tuning according to the evaluation result, such as adjusting super parameters, increasing training iteration times and the like, so as to improve the delimitation accuracy and generalization capability of the model.
It should be noted that the transducer model is a neural network based on the attention mechanism, and the structures of the CNN (Convolutional Neural Network ) and the RNN (Recurrent Neural Network, recurrent neural network) are discarded. As an example of an encoder-decoder architecture, the encoder and decoder are stacked based on self-attention modules, and embedding (ebedding) of the source (input) sequence and the target (output) sequence means that position coding (positional encoding) is added to the source (input) sequence and the target (output) sequence, respectively, and the position coding (positional encoding) is input to the encoder and the decoder, respectively, to obtain an effective prediction output.
The transducer encoder is built up of a plurality of identical layers, each layer having two sub-layers (sub-layers are denoted sublayers). The first sub-layer is a multi-head self-attention (multi-head self-attention) convergence; the second sub-layer is a location based feed forward network (positionwise feed-forward network). In calculating the self-attention of the encoder, the query, key and value are all from the output of the previous encoder layer.
The transform decoder is also built up from multiple identical layers, and residual connections and layer normalization are used in the layers. In addition to the two sublayers described in the encoder, the decoder inserts a third sublayer, called encoder-decoder attention (encoder-decoder attention) layer, between the two sublayers. In encoder-decoder attention, the output from the previous decoder layer is queried, while the keys and values are from the output of the entire encoder. In decoder self-attention, the query, key and value are all from the output of the last decoder layer.
According to the method and the device, the network problem can be automatically delimited by training the delimiting model, so that the time and the workload of manual positioning are reduced, and automatic, efficient and accurate problem positioning is realized.
Based on the above embodiment, before the acquiring the positioning data and the delimiting data based on the network data of the at least one data source, the method further includes:
step 110, constructing a warehouse entry table;
step 120, configuring a data processing task based on the data docking mode, the information of the data source to which the data belongs, the task execution information and the task flow chart;
and 130, accessing network data of at least one data source to a data management system based on the warehouse entry table and the data processing task.
To simplify data management, enriching positioning and delimiting data, network data dispersed in different systems or data sources may be managed into one system.
Specifically, a warehouse entry table and a configuration data processing task (namely an ETL task) are constructed, and then multi-source data are accessed from different systems into a home system (namely a data management system) of global quality end-to-end positioning delimitation capacity based on the warehouse entry table and the data processing task, so that support capacity construction is facilitated. The ETL (Extract, transform, load) task needs to configure four kinds of information, which are respectively: data docking modes, such as a file SFTP mode and a database mode; for the related information of the data source of the data of different data docking modes, such as server IP, storage path, account number password, file name, etc. of the file SFTP (Secure File Transfer Protocol ) mode, database IP, port number, account number password, data table name, etc. of the database mode; task execution information such as execution time, execution cycle, execution frequency, etc.; and a task flow chart, for example, presenting all links of the ETL task in a flow chart mode and supporting front-end configuration.
For example, based on the binning table and ETL tasks, the steps of accessing multi-source data from different systems to the home system for global quality end-to-end location delimitation capability are as follows:
data source analysis: the type, structure and source of the multi-source data are known in detail. For example, the data source that needs to be accessed, including its format, protocol, access manner, etc., is determined.
Designing a warehouse entry table: appropriate database table structures are designed to store the accessed data according to the requirements. For example, considering the need for global quality end-to-end positioning delimitation, the table structure should contain enough fields to store the required information, such as time stamps, key performance indicators, alarm information, etc.
Creating a database table: according to the design, a corresponding database table is created in the home system, and the table structure is ensured to be consistent with the design.
ETL task configuration: appropriate ETL tools or methods are selected to effect extraction, conversion and loading of data. According to the characteristics of the data source and the requirements of the attribution system, the ETL task is configured to extract the source data from different systems, the source data are converted and cleaned, and finally the data are loaded into a warehouse entry table of the attribution system.
Data synchronization and scheduling: reasonable data synchronization frequency and scheduling strategy are set, and real-time or timeliness of data from a source system to a home system is ensured.
And (3) data quality monitoring: and establishing a data quality monitoring mechanism, and periodically checking and evaluating the accuracy, the integrity and the consistency of the accessed data to ensure the data quality.
Security and rights control: in the data access process, the security and privacy of the data are ensured, and proper authority control is set to prevent unauthorized access and data disclosure.
Testing and verifying: after the ETL task is configured, testing and verification are performed, so that data is ensured to be correctly accessed into a warehouse entry table of the home system from the multi-source system, and the construction of the global quality end-to-end positioning delimitation capability can be supported.
Further, processing is carried out on the accessed multi-source data to form sample data for supporting model training and parameter entering data for capability calling, and global quality end-to-end positioning delimitation capability construction based on an AI algorithm is realized. Specifically, extracting performance indexes of the positioning data, and associating the performance indexes with the network objects to generate a data table with the performance indexes associated with the network objects; root cause data of the delimited data is extracted, and the root cause data is associated with a solution library to generate delimited conclusion data and solutions. For example, the positioning data and the delimitation data are processed respectively, and the processing procedure is as follows:
positioning data processing: and extracting key performance indexes with strong correlation of quality from the signaling performance data and the OMC northbound performance data, correlating the two types of performance data with different objects to form a wide table of correlated indexes of the sub-objects, wherein the key performance indexes comprise indexes related to network connection quality, response time, packet loss rate and the like.
And (3) delimitation data processing: signaling error code data, associating an error code solution library, and generating error conclusion and solutions; alarm data, extracting alarm reasons, associating alarm solution libraries, and generating alarm conclusion and solutions; and the log data is used for extracting error code information in the log, associating an error code solution library and generating an error conclusion and a solution.
According to the embodiment of the application, the network data dispersed in different systems or data sources are managed in one system, so that the data management process is simplified, the dispersibility of the data is reduced, the consistency and the integrity of the data are improved, the data are conveniently comprehensively analyzed and utilized, and a more comprehensive and accurate data basis is provided for capacity construction. Meanwhile, the positioning data and the delimitation data are processed to obtain sample data supporting model training and parameter entering data for capability calling, so that support is provided for positioning and delimitation of network quality problems, and the accuracy of positioning delimitation of the network quality problems is improved.
For further explanation of the network quality problem delimiting method proposed in the present application, refer to fig. 2 and the following embodiments.
According to the embodiment of the application, the positioning and delimitation of the quality problem of the intelligent network are realized through the encapsulated data access module, the encapsulated data processing module, the encapsulated data labeling module, the end-to-end quality positioning module and the end-to-end quality delimitation module. Wherein, the functions of each module are as follows:
and the data access module is used for: the access nano tube mainly realizes multi-source data and comprises signaling data, OMC northbound data, alarm data, dial testing data, log data, error code solution library and alarm solution library. Specifically, a warehouse entry table is built, ETL tasks are configured, and then multi-source data is uniformly accessed into a home system of global quality end-to-end positioning delimitation capacity from different systems based on the warehouse entry table and the ETL tasks, so that the construction of supporting capacity is facilitated.
And a data processing module: the method mainly processes the accessed multi-source data to form sample data for supporting model training and parameter entering data for capability calling, and realizes the construction of the global quality end-to-end positioning delimitation capability based on an AI algorithm. Specifically including processing of positioning data and bounding data.
The data marking module: the quality difference labeling of the whole domain object is mainly realized, wherein the quality difference labeling can be realized by a supervised AI algorithm supporting the end-to-end positioning capability of the quality, and model training data with quality difference samples and non-quality difference samples are formed through the labeled quality difference object; and meanwhile, the marked global quality difference object is used as a trigger, the end-to-end delimitation capability of the quality is called, the root cause positioning of the quality difference object is realized, a problem solution is provided, and the quality problem is guided to be repaired. The labeling mode comprises expert experience labeling and AI algorithm labeling, wherein the expert experience labeling is mainly based on quality difference object labeling from the performance index level by comparing key performance index threshold values and index same ring ratio fluctuation threshold values, and simultaneously, alarming objects, dial test failure objects and log error objects are marked as quality difference objects, so that a full-domain problem object library of a primary edition is constructed. The AI algorithm labeling is mainly based on multi-source data of performance, alarm, dial test and log, and is introduced to learn, and labeling objects are poor in quality.
Mass end-to-end positioning module: based on manual or preset time self-triggering, the AI algorithm is invoked by inputting the performance index data of the global object, so as to realize quality difference positioning, and the positioning result (quality difference time, quality difference object and quality difference index) is output. The model training involves three links of training sample preparation, sample data preprocessing, algorithm selection and training tuning.
Mass end-to-end delimitation module: and triggering and starting by a quality end-to-end positioning module, and realizing the output of the delimitation conclusion by calling an AI algorithm through the input capacity input parameter. The capacity input parameter is divided into two parts, namely a delimitation object which is the output of a quality end-to-end positioning module, namely quality difference time, quality difference objects and quality difference indexes; secondly, error code data, alarm data, dial test data and log data which assist in carrying out delimitation conclusion output are generated by associating the two types of input parameters into multi-dimensional potential cause data of a quality difference object, contribution ratio calculation is carried out by using an AI algorithm, TOPN quality difference causes are output, and problem repair is guided.
According to the embodiment of the application, the problem of global quality of positioning and delimiting of expert experience and an AI algorithm is fused, the problem that the existing static threshold mode cannot meet the requirements of flexibility and practicability of a complex network is solved, and the accuracy of the problem of positioning and delimiting of the network quality is improved.
Fig. 3 is a schematic structural diagram of a network quality problem delimiting device provided in the present application, and referring to fig. 3, an embodiment of the present application provides a network quality problem delimiting device, which includes a data acquisition module 301, a positioning module 302, and a delimiting module 303.
A data acquisition module 301, configured to acquire positioning data and delimitation data based on network data of at least one data source;
the positioning module 302 is configured to input the positioning data into a positioning model, and obtain positioning result data output by the positioning model; the positioning model is obtained by training sample positioning data;
the delimiting module 303 is configured to input the positioning result data and the delimiting data into a delimiting model if it is determined that a network quality difference problem exists based on the positioning result data, and obtain quality difference cause data output by the delimiting model; the delimitation model is obtained by training sample positioning result data and sample delimitation data.
According to the network quality problem delimiting device, positioning data and delimiting data are obtained through network data based on at least one data source; inputting the positioning data into a positioning model to obtain positioning result data output by the positioning model; the positioning model is obtained by training sample positioning data; if the network quality difference problem exists based on the positioning result data, inputting the positioning result data and the delimitation data into a delimitation model, and obtaining quality difference reason data output by the delimitation model; the delimitation model is obtained by training sample positioning result data and sample delimitation data. According to the embodiment of the application, the intelligent recognition, positioning and delimitation of the network quality problem are realized by adopting the positioning model and the delimitation model, a large amount of network data can be analyzed more quickly and efficiently by the model, the problem is accurately positioned, the manual investigation time is reduced, and the accuracy and the efficiency of the positioning and delimitation of the network quality problem are improved.
In one embodiment, the network quality problem delimiting means further comprises a training module for:
collecting the sample positioning data;
performing data annotation on the sample positioning data to obtain training data;
preprocessing the training data, wherein the preprocessing comprises missing value processing and outlier processing;
and training the preprocessed training data by adopting a first target algorithm to obtain the positioning model.
In one embodiment, the training module is further configured to:
obtaining a quality difference sample and a non-quality difference sample of the sample positioning data;
based on a set labeling strategy or a labeling model, carrying out data labeling on the quality difference sample;
and determining the training data based on the marked quality difference sample and the marked non-quality difference sample.
In one embodiment, the training module is further configured to:
collecting the sample positioning result data and the sample delimitation data;
and training the sample positioning result data and the sample delimitation data by adopting a second target algorithm to obtain the delimitation model.
In one embodiment, the data acquisition module 301 is further configured to:
constructing a warehouse entry table;
configuring a data processing task based on a data docking mode, information of a data source to which data belongs, task execution information and a task flow chart;
and accessing network data of at least one data source to a data management system based on the warehouse entry table and the data processing task.
In one embodiment, the data acquisition module 301 is further configured to:
extracting performance indexes of the positioning data, and associating the performance indexes with network objects to generate a data table with the performance indexes associated with the network objects;
root cause data of the delimited data is extracted, and the root cause data is associated with a solution library to generate delimited conclusion data and a solution.
In one embodiment, the delimiting module 303 is further configured to:
acquiring quality difference time, quality difference objects and quality difference indexes based on the positioning result data;
and determining that a network quality difference problem exists based on the quality difference time, the quality difference object and the quality difference index.
Fig. 4 illustrates a physical schematic diagram of an electronic device, as shown in fig. 4, which may include: processor 410, communication interface 420, memory 430 and communication bus 440, wherein processor 410, communication interface 420 and memory 430 communicate with each other through communication bus 440. The processor 410 may invoke logic instructions in the memory 430 to perform a network quality problem delimitation method comprising:
acquiring positioning data and delimitation data based on network data of at least one data source;
inputting the positioning data into a positioning model to obtain positioning result data output by the positioning model; the positioning model is obtained by training sample positioning data;
if the network quality difference problem exists based on the positioning result data, inputting the positioning result data and the delimitation data into a delimitation model, and obtaining quality difference reason data output by the delimitation model; the delimitation model is obtained by training sample positioning result data and sample delimitation data.
Further, the logic instructions in the memory 430 described above may be implemented in the form of software functional units and may be stored in a computer-readable storage medium when sold or used as a stand-alone product. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: a usb disk, a removable hard disk, a Read-only memory (ROM), a random access memory (RAM, randomAccessMemory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
In another aspect, the present application also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, is implemented to perform a network quality problem delimiting method provided by the above methods, the method comprising:
acquiring positioning data and delimitation data based on network data of at least one data source;
inputting the positioning data into a positioning model to obtain positioning result data output by the positioning model; the positioning model is obtained by training sample positioning data;
if the network quality difference problem exists based on the positioning result data, inputting the positioning result data and the delimitation data into a delimitation model, and obtaining quality difference reason data output by the delimitation model; the delimitation model is obtained by training sample positioning result data and sample delimitation data.
The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present application, and are not limiting thereof; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the corresponding technical solutions.
Claims (10)
1. A method for delimiting a network quality problem, comprising:
acquiring positioning data and delimitation data based on network data of at least one data source;
inputting the positioning data into a positioning model to obtain positioning result data output by the positioning model; the positioning model is obtained by training sample positioning data;
if the network quality difference problem exists based on the positioning result data, inputting the positioning result data and the delimitation data into a delimitation model, and obtaining quality difference reason data output by the delimitation model; the delimitation model is obtained by training sample positioning result data and sample delimitation data.
2. The network quality problem delimitation method according to claim 1, characterized in that the positioning model is trained based on the following steps:
collecting the sample positioning data;
performing data annotation on the sample positioning data to obtain training data;
preprocessing the training data, wherein the preprocessing comprises missing value processing and outlier processing;
and training the preprocessed training data by adopting a first target algorithm to obtain the positioning model.
3. The method for delimiting network quality problems according to claim 2, wherein the performing data labeling on the sample positioning data to obtain training data includes:
obtaining a quality difference sample and a non-quality difference sample of the sample positioning data;
based on a set labeling strategy or a labeling model, carrying out data labeling on the quality difference sample;
and determining the training data based on the marked quality difference sample and the marked non-quality difference sample.
4. The network quality problem delimitation method according to claim 1, characterized in that the delimitation model is trained based on the following steps:
collecting the sample positioning result data and the sample delimitation data;
and training the sample positioning result data and the sample delimitation data by adopting a second target algorithm to obtain the delimitation model.
5. The method for delimiting a network quality problem according to claim 1, wherein before the acquiring of the positioning data and the delimiting data based on the network data of the at least one data source, further comprises:
constructing a warehouse entry table;
configuring a data processing task based on a data docking mode, information of a data source to which data belongs, task execution information and a task flow chart;
and accessing network data of at least one data source to a data management system based on the warehouse entry table and the data processing task.
6. The method for delimiting a network quality problem according to claim 5, wherein after the network data of the at least one data source is accessed to the data management system based on the binning table and the data processing task, further comprising:
extracting performance indexes of the positioning data, and associating the performance indexes with network objects to generate a data table with the performance indexes associated with the network objects;
root cause data of the delimited data is extracted, and the root cause data is associated with a solution library to generate delimited conclusion data and a solution.
7. The network quality problem delimitation method according to claim 1, characterized in that determining that a network quality difference problem exists based on the positioning result data comprises:
acquiring quality difference time, quality difference objects and quality difference indexes based on the positioning result data;
and determining that a network quality difference problem exists based on the quality difference time, the quality difference object and the quality difference index.
8. A network quality problem delimiter, comprising:
the data acquisition module is used for acquiring positioning data and delimitation data based on network data of at least one data source;
the positioning module is used for inputting the positioning data into a positioning model and obtaining positioning result data output by the positioning model; the positioning model is obtained by training sample positioning data;
the delimiting module is used for inputting the positioning result data and the delimiting data into a delimiting model if the network quality difference problem exists based on the positioning result data, and obtaining quality difference cause data output by the delimiting model; the delimitation model is obtained by training sample positioning result data and sample delimitation data.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the network quality problem delimitation method according to any of claims 1 to 7 when executing the program.
10. A non-transitory computer readable storage medium, having stored thereon a computer program, which when executed by a processor implements the network quality problem delimitation method according to any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311090496.1A CN117376087A (en) | 2023-08-28 | 2023-08-28 | Method, device, equipment and storage medium for delimiting network quality problems |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311090496.1A CN117376087A (en) | 2023-08-28 | 2023-08-28 | Method, device, equipment and storage medium for delimiting network quality problems |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117376087A true CN117376087A (en) | 2024-01-09 |
Family
ID=89399054
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311090496.1A Pending CN117376087A (en) | 2023-08-28 | 2023-08-28 | Method, device, equipment and storage medium for delimiting network quality problems |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117376087A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118055427A (en) * | 2024-04-16 | 2024-05-17 | 中国电信股份有限公司浙江分公司 | Method and device for automatic network optimization of private network base station |
-
2023
- 2023-08-28 CN CN202311090496.1A patent/CN117376087A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118055427A (en) * | 2024-04-16 | 2024-05-17 | 中国电信股份有限公司浙江分公司 | Method and device for automatic network optimization of private network base station |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108415789B (en) | Node fault prediction system and method for large-scale hybrid heterogeneous storage system | |
| CN109255440B (en) | A Recurrent Neural Network (RNN)-Based Method for Predictive Maintenance of Power Generation Equipment | |
| US8019584B2 (en) | Method and system for modeling likely invariants in distributed systems | |
| CN109471698B (en) | System and method for detecting abnormal behavior of virtual machine in cloud environment | |
| CN117669176A (en) | Power distribution network reliability assessment method and device based on network simplification method | |
| CN113891342A (en) | Base station inspection method, device, electronic device and storage medium | |
| CN116167370A (en) | Anomaly Detection Method of Distributed System Based on Log Spatiotemporal Feature Analysis | |
| CN117376087A (en) | Method, device, equipment and storage medium for delimiting network quality problems | |
| CN119728397B (en) | Network fault prediction method and system | |
| CN119044343A (en) | Method for establishing gas model on oil surface under sudden severe fault of oil immersed transformer | |
| CN120197957A (en) | An abnormal behavior analysis system for data centers based on artificial intelligence | |
| CN120277571A (en) | Construction method of multidimensional time sequence anomaly detection system | |
| CN120034421A (en) | A data processing method based on intelligent operation and maintenance management cloud platform | |
| CN118861950B (en) | Transformer multi-parameter fusion state evaluation method, device, equipment and medium | |
| CN119669883A (en) | Semiconductor automated testing method and testing equipment based on artificial intelligence | |
| CN119248617A (en) | Elasticsearch cluster fault intelligent detection system | |
| CN118041692B (en) | Network security testing method and system based on intrusion detection technology | |
| CN110263811A (en) | A kind of equipment running status monitoring method and system based on data fusion | |
| CN118130951A (en) | Railway power load dynamic mutation-based monitoring method | |
| Li et al. | Anomaly detection of internet service quality degradation in digital twin for fixed access network | |
| CN110060374B (en) | Method and device for detecting abnormality of aircraft fuel system | |
| CN106487592A (en) | A kind of Faults in Distributed Systems diagnostic method based on data cube | |
| Bai | Network equipment fault maintenance decision system based on bayesian decision algorithm | |
| CN118971381B (en) | Method and device for sensing power grid situation and managing faults based on generated AI | |
| CN119065936B (en) | A cloud platform-related microservice fault warning method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |