WO2020108159A1 - 一种网络故障根因检测方法、系统及存储介质 - Google Patents

一种网络故障根因检测方法、系统及存储介质 Download PDF

Info

Publication number
WO2020108159A1
WO2020108159A1 PCT/CN2019/111750 CN2019111750W WO2020108159A1 WO 2020108159 A1 WO2020108159 A1 WO 2020108159A1 CN 2019111750 W CN2019111750 W CN 2019111750W WO 2020108159 A1 WO2020108159 A1 WO 2020108159A1
Authority
WO
WIPO (PCT)
Prior art keywords
sample
root cause
model
diagnosed
network
Prior art date
Application number
PCT/CN2019/111750
Other languages
English (en)
French (fr)
Inventor
杨正华
Original Assignee
中兴通讯股份有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 中兴通讯股份有限公司 filed Critical 中兴通讯股份有限公司
Publication of WO2020108159A1 publication Critical patent/WO2020108159A1/zh

Links

Images

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
    • H04L41/14—Network analysis or design
    • H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
    • 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
    • H04L41/0677—Localisation of faults
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/06—Testing, supervising or monitoring using simulated traffic

Definitions

  • the present invention requires the priority of the Chinese patent application filed on November 26, 2018 in the Chinese Patent Office with the application number 201811420214.9 and the invention titled "A network failure root cause detection method, system and storage medium”. The content is incorporated by reference in the present invention.
  • the invention relates to the technical field of mobile communication networks, in particular to a method, system and storage medium for detecting the root cause of network failures.
  • the traditional wireless communication network problem location mainly adopts manual analysis, specifically network alarm and configuration data analysis, and gradually investigates the cause of the problem to determine the root cause of the failure.
  • the efficiency of this traditional fault root cause location method is very low, and the accuracy of fault root cause location relies heavily on the knowledge, experience, and skills of engineers, and does not adapt to current network development trends.
  • At least one embodiment of the present invention provides a method, system, and storage medium for detecting the root cause of a network failure.
  • an embodiment of the present invention provides a network failure root cause detection method.
  • the detection method includes: acquiring multiple network performance parameters, processing each of the network performance parameters separately to obtain model samples; Perform fine-grained clustering on the model samples, map model samples with a similarity greater than a preset threshold to the same sample unit, and obtain multiple sample units; add a root cause label to each sample unit to generate a root cause detection model; to be diagnosed Network performance parameters are processed to obtain samples to be diagnosed, and the similarity value of the sample to be diagnosed and each of the model samples is calculated separately; the sample unit corresponding to the model sample with the highest similarity to the sample to be diagnosed is obtained Root cause label, using the root cause label as the root cause of the failure of the sample to be diagnosed.
  • the embodiments of the present invention may also make the following improvements.
  • adding a root cause label to each sample unit to generate a root cause detection model includes: sorting the sample units according to the number of model samples in the sample unit ; Set the sample units ranked above the preset threshold as the core unit, and use the density-based clustering algorithm to merge each core unit with other sample units to obtain multiple sample unit clusters; for each sample unit cluster Add a root cause label to generate a root cause detection model.
  • the performing fine-grained clustering on all the model samples to map model samples with a similarity greater than a preset threshold to the same sample unit includes: All the model samples are input to the input layer of the self-organizing mapping neural network; the fine-grained clustering of all the model samples is completed through the self-organizing mapping neural network competition layer, and the model samples with similarity greater than a preset threshold are mapped to the same In the sample unit.
  • the processing each of the network performance parameters separately to obtain a model sample includes: using a principal component analysis method for each of the network performance parameters Perform dimensionality reduction processing to obtain the model sample.
  • the method before performing dimensionality reduction processing on each of the network performance parameters separately, the method further includes: Perform filtering to remove outliers in the network performance parameters.
  • the detection method further includes: using the model samples as reference model samples; and performing data standardization processing on all reference model samples separately To get the model sample.
  • the processing of the network performance parameter to be diagnosed to obtain a sample to be diagnosed includes: dimensionality reduction of the network performance parameter to be diagnosed by a principal component analysis method Processing to obtain the sample to be diagnosed.
  • the Calculating the similarity value between the sample to be diagnosed and each of the model samples includes separately calculating the Euclidean distance between the sample to be diagnosed and each of the model samples as the similarity value.
  • an embodiment of the present invention provides a network failure root cause detection system, the network failure root cause detection system includes a processor and a memory; the processor is used to execute a failure root cause detection program stored in the memory To implement the fault root cause detection method described in any one of the embodiments of the first aspect.
  • an embodiment of the present invention provides a computer-storable medium that stores one or more programs, and the one or more programs can be executed by one or more processors to implement The fault root cause detection method described in any one of the embodiments of the first aspect.
  • FIG. 1 is a schematic flowchart of a method for detecting a root cause of a network failure according to an embodiment of the present invention
  • FIG. 2 is a schematic flowchart of a network fault root cause detection method according to another embodiment of the present invention.
  • FIG. 3 is a schematic structural diagram of a network failure root cause detection system according to another embodiment of the present invention.
  • a method for detecting the root cause of a network failure includes: S11. Obtain multiple network performance parameters, and process each network performance parameter separately to obtain a model sample.
  • a large number of wireless communication network performance parameters such as communication quality, connection duration, communication range, and other network performance parameters are collected, and the network performance parameters are processed to obtain Model samples belonging to each cell.
  • the network performance parameters of each cell can be filtered to remove outliers in the network performance parameters.
  • the filtering method can filter outliers in your network performance parameters by the Laida criterion, and pass the filtered network performance parameters.
  • Principal component analysis method is used for dimensionality reduction, and the network performance parameters after dimensionality reduction are subjected to data standardization processing, such as normalization method, normalization method, and normalization method.
  • the normalization of data is to scale the data to make it fall. Into a small specific interval. It is often used in the processing of some comparison and evaluation indicators to remove the unit limit of the data and convert it into a dimensionless pure value, and finally obtain the above model sample.
  • clustering the process of dividing the collection of physical or abstract objects into multiple classes composed of similar objects by fine-grained clustering of all model samples.
  • the cluster generated by clustering is a set of data objects. These objects are similar to the objects in the same cluster and different from the objects in other clusters.
  • cluster analysis content is very rich, including systematic clustering method, ordered sample clustering method, dynamic clustering method, fuzzy clustering method, graph theory clustering method, cluster prediction method Wait.
  • the clustering of different model samples is completed, so that similar model samples are clustered into the same class, that is, the network performance parameters of cells with the same root cause are clustered into the same class.
  • the model samples are processed through the self-organizing map neural network, and all model samples are input into the input layer of the self-organizing map neural network, and the fine-grained aggregation of all model samples is completed through the competition layer of the self-organizing map neural network.
  • Class, mapping model samples with similarity greater than a preset threshold to the same sample unit, self-organizing mapping neural network is a competitive learning unsupervised neural network, which can map high-dimensional spatial data to low-dimensional topology while maintaining topological structure Space, to perform visual cluster analysis.
  • the self-organizing mapping neural network is composed of an input layer and a competition layer.
  • the input layer is composed of N neurons, and N corresponds to the number of model samples;
  • the competition layer is generally designed in a two-dimensional array, and each neuron in the competition layer is related to the output layer. Neurons are connected, and the weight of each neuron represents a type of pattern in the input space.
  • the winning neuron is adjusted to keep approaching the input mode, and the weights of the neurons surrounding the winning neuron are updated to maintain the topological structure of the input space.
  • the model samples processed through the above steps have model samples with similar symptoms in the same sample unit, and the symptom characteristics of the model samples in the sample unit adjacent to the sample unit are also relatively similar.
  • model samples in each sample unit can be combined based on expert knowledge and experience, and a root cause label can be added to each sample unit, or the parameters of the model sample in the sample unit can be combined based on past processing experience. , Add a root cause label to the sample unit.
  • S14 Process the performance parameters of the network to be diagnosed to obtain a sample to be diagnosed, and calculate the similarity value of the sample to be diagnosed and each model sample separately.
  • the network performance parameter to be diagnosed obtains the sample to be diagnosed by the same processing method in the above steps, and respectively calculates the similarity value of the sample to be diagnosed and each model sample in the sample unit.
  • the distance between the sample to be diagnosed and the model sample is calculated as the similarity value between the sample to be diagnosed and the model sample.
  • the Euclidean distance between the sample to be diagnosed and the model sample can be calculated as the sample to be diagnosed and The similarity value of model samples, the smaller the Euclidean distance value, the higher the similarity value.
  • the root cause label of the sample unit of the model sample with the highest similarity to the sample to be diagnosed is taken as the root cause of the failure of the sample to be diagnosed.
  • an embodiment of the present invention also provides a network failure root cause detection method. Compared with the detection method shown in FIG. 1, the difference is that a root cause label is added to each sample unit to generate a root cause detection model , Including: S21, sort the sample units according to the number of model samples in the sample unit.
  • the sample units are sorted according to the number of model samples in the sample unit, and the sample unit with the larger number of model samples is ranked higher.
  • sample units ranked above the preset threshold are set as core units, and the other sample units are merged with each core unit separately, so that sample units with similar symptom phenomena are combined into sample unit clusters.
  • model samples in each sample unit cluster may be combined based on expert knowledge and experience, and a root cause label may be added to each sample unit cluster, or the model samples in the sample unit may be combined based on past processing experience. Add the root cause labels to the sample unit clusters.
  • an embodiment of the present invention provides a network failure root cause detection system.
  • the network failure root cause detection system includes a processor and a memory; the processor is used to execute a failure root cause detection program stored in the memory to implement the first
  • the fault root cause detection method of any embodiment includes a processor and a memory; the processor is used to execute a failure root cause detection program stored in the memory to implement the first
  • a storage medium for recording the program code of the software program that can realize the functions of the above-mentioned embodiment to the system or device in the above embodiment, and read and execute the storage in the storage by the computer (or CPU or MPU) of the system or device The program code in the media.
  • the program code itself read from the storage medium performs the functions of the above-described embodiments, and the storage medium storing the program code constitutes an embodiment of the present invention.
  • a storage medium for providing program codes for example, a floppy disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a magnetic tape, a nonvolatile memory card, a ROM, and the like can be used.
  • the functions of the above embodiments can be realized not only by the computer executing the read program code, but also by some or all of the actual processing operations performed by the OS (operating system) running on the computer according to the instructions of the program code.
  • the embodiments of the present invention also include a case where the program code read from the storage medium is written into the function expansion card inserted into the computer, or is written into the memory provided in the function expansion unit connected to the computer After that, the CPU or the like included in the function expansion card or the function expansion unit performs part or all of the processing in accordance with the instructions of the program code, thereby realizing the functions of the above-described embodiments.
  • An embodiment of the present invention provides a computer storable medium, which stores one or more programs, and the one or more programs can be executed by one or more processors to implement any of the first aspect Examples of fault root cause detection methods.
  • the above technical solution of the present invention has the following advantages:
  • the model samples generated by the network performance parameters are mapped to the same sample unit through clustering, and the sample units are separately Root cause tags are added to generate a root cause detection model, and the network performance parameters to be diagnosed are diagnosed through the root cause detection model, so as to quickly diagnose the root cause of a fault in a wireless communication network and improve the efficiency of fault diagnosis.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

本发明涉及一种网络故障根因检测方法、系统及存储介质，检测方法包括：分别对每个网络性能参数进行处理，得到模型样本；将相似度大于预设阈值的模型样本映射到同一样本单元；对每个样本单元添加根因标签；分别计算待诊断样本与每个模型样本的相似度值；获取与待诊断样本相似度最高的模型样本对应的根因标签作为苏搜狐待诊断样本的故障根因。

Description

一种网络故障根因检测方法、系统及存储介质
交叉引用
本发明要求在2018年11月26日提交中国专利局、申请号为201811420214.9、发明名称为“一种网络故障根因检测方法、系统及存储介质”的中国专利申请的优先权，该申请的全部内容通过引用结合在本发明中。
技术领域
本发明涉及移动通信网络技术领域，尤其涉及一种网络故障根因检测方法、系统及存储介质。
背景技术
随着人们对移动通信业务需求的日益增长，移动网络运营规模越来越庞大，所采用的技术也越来越复杂。因而网络发生故障的频率越来越高，定位难度也不断增大，最终导致网络运营维护成本急剧增长。
传统的无线通信网络问题根因定位主要采取人工分析方式，具体而言网络警和配置数据进行分析，对问题原因逐步排查后确定故障根因。这种传统的故障根因定位方法的效率非常低下，而且故障根因定位准确率严重依赖工程师的知识、经验和技能，不适应当前网络的发展趋势。
发明内容
为了解决现有技术存在的问题，本发明的至少一个实施例提供了一种网络故障根因检测方法、系统及存储介质。
第一方面，本发明实施例提供了一种网络故障根因检测方法，所述检测方法包括：获取多个网络性能参数，分别对每个所述网络性能参数进行处理，得到模型样本；对所有所述模型样本进行细粒度聚类，将相似度大于预设阈值的模型样本映射到同一样本单元，得到多个样本单元；对每个样本单元添 加根因标签，生成根因检测模型；对待诊断网络性能参数进行处理，得到待诊断样本，分别计算所述待诊断样本与每个所述模型样本的相似度值；获取与所述待诊断样本相似度最高的所述模型样本对应的样本单元的根因标签，将所述根因标签作为所述待诊断样本的故障根因。
基于上述技术方案，本发明实施例还可以做出如下改进。
结合第一方面，在第一方面的第一种实施例中，所述对每个样本单元添加根因标签，生成根因检测模型，包括：根据样本单元中模型样本的数量对样本单元进行排序；将排序高于预设阈值的样本单元设置为核心单元，采用基于密度的聚类算法将每个核心单元与其他样本单元进行合并，得到多个样本单元类簇；对每个样本单元类簇添加根因标签，生成根因检测模型。
结合第一方面，在第一方面的第二种实施例中，所述对所有所述模型样本进行细粒度聚类，将相似度大于预设阈值的模型样本映射到同一样本单元，包括：将所有所述模型样本输入自组织映射神经网络的输入层；通过自组织映射神经网络的竞争层完成对所有所述模型样本的细粒度聚类，将相似度大于预设阈值的模型样本映射到同一所述样本单元中。
结合第一方面，在第一方面的第三种实施例中，所述分别对每个所述网络性能参数进行处理，得到模型样本，包括：通过主成分分析方法对每个所述网络性能参数进行降维处理，得到所述模型样本。
结合第一方面的第三种实施例，在第一方面的第四种实施例中，所述分别对每个所述网络性能参数进行降维处理之前，还包括：对所有所述网络性能参数进行过滤，剔除所述网络性能参数中的异常值。
结合第一方面的第三种实施例，在第一方面的第五种实施例中，所述检测方法还包括：将所述模型样本作为参考模型样本；对所有参考模型样本分别进行数据标准化处理，得到所述模型样本。
结合第一方面，在第一方面的第六种实施例中，所述对待诊断网络性能参数进行处理，得到待诊断样本，包括：通过主成分分析方法对所述待诊断 网络性能参数进行降维处理，得到所述待诊断样本。
结合第一方面或第一方面的第一、第二、第三、第四、第五或第六种实施例中任一种实施例，在第一方面的第七种实施例中，所述计算所述待诊断样本与每个所述模型样本的相似度值，包括：分别计算待诊断样本与每个所述模型样本的欧氏距离，作为所述相似度值。
第二方面，本发明实施例提供了一种网络故障根因检测系统，所述网络故障根因检测系统包括处理器、存储器；所述处理器用于执行所述存储器中存储的故障根因检测程序，以实现第一方面中任一项实施例所述的故障根因检测方法。
第三方面，本发明实施例提供了一种计算机可存储介质，所述计算机可存储介质存储有一个或者多个程序，所述一个或者多个程序可被一个或者多个处理器执行，以实现第一方面中任一项实施例所述的故障根因检测方法。
附图说明
图1是本发明实施例提供的一种网络故障根因检测方法流程示意图；
图2是本发明另一实施例提供的一种网络故障根因检测方法流程示意图；
图3是本发明又一实施例提供的一种网络故障根因检测系统结构示意图。
具体实施方式
为使本发明实施例的目的、技术方案和优点更加清楚，下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本发明的一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动的前提下所获得的所有其他实施例，都属于本发明保护的范围。
如图1所示，本发明实施例提供的一种网络故障根因检测方法，检测方法包括：S11、获取多个网络性能参数，分别对每个网络性能参数进行处理， 得到模型样本。
在本实施例中，通过对无线通信网络中各个小区进行数据采集，收集大量无线通信网络性能参数，比如，通信质量、连接时长、通信范围等网络性能参数，通过对网络性能参数进行处理，得到属于各个小区的模型样本。
比如，可以通过对各个小区的网络性能参数进行过滤，剔除网络性能参数中的异常值，过滤方法可以通过拉依达准则贵网络性能参数中的异常值进行过滤，对过滤后的网络性能参数通过主成分分析方法进行降维，对降维后的网络性能参数进行数据标准化处理，如规范化方法、正规化方法和归一化方法，数据的标准化(normalization)是将数据按比例缩放，使之落入一个小的特定区间。在某些比较和评价的指标处理中经常会用到，去除数据的单位限制，将其转化为无量纲的纯数值，最终得到上述模型样本。
S12、对所有模型样本进行细粒度聚类，将相似度大于预设阈值的模型样本映射到同一样本单元，得到多个样本单元。
在本实施例中，可通过将所有模型样本进行细粒度聚类，将物理或抽象对象的集合分成由类似的对象组成的多个类的过程被称为聚类。由聚类所生成的簇是一组数据对象的集合，这些对象与同一个簇中的对象彼此相似，与其他簇中的对象相异。通过对模型样本进行细粒度聚类，聚类分析内容非常丰富，有系统聚类法、有序样品聚类法、动态聚类法、模糊聚类法、图论聚类法、聚类预报法等。完成了对不同模型样本的聚类，使得将相似的模型样本聚到同一类中，即具有相同根因的小区的网络性能参数聚到同一类中。
本实施例中，通过自组织映射神经网络对模型样本进行处理，将所有模型样本输入自组织映射神经网络的输入层中，通过自组织映射神经网络的竞争层完成对所有模型样本的细粒度聚类，将相似度大于预设阈值的模型样本映射到同一样本单元中，自组织映射神经网络是一种竞争学习型的无监督神经网络，能将高维空间数据保持拓扑结构地映射到低维空间，进行可视化的聚类分析。自组织映射神经网络由输入层和竞争层组成，输入层由N个神经 元组成，N对应模型样本的数量；竞争层一般采用二维阵列式设计，竞争层中每个神经元都与输出层神经元相连接，每个神经元的权值代表了输入空间的一类模式。自组织映射神经网络学习过程中调整获胜神经元使之不断向输入模式靠拢，同时更新获胜神经元周围神经元的权值，以保持输入空间的拓扑结构。因此样本空间经过自组织映射神经网络映射后可得到它的分布骨架，或者说完成了一次细粒度聚类，把相似度大的样本映射到同一个神经元，即该神经元代表了样本空间中若干高度相似的样本子集。通过上述步骤处理后的模型样本，同一样本单元中具有若干相似症状的模型样本，而与该样本单元相邻的样本单元中的模型样本的症状特征也较为相近。
S13、对每个样本单元添加根因标签，生成根因检测模型。
通过对每个样本单元添加根因标签，即可完成对该样本单元下的所有模型样本的故障根因的确认，提高对模型样本的故障根因的确定速率，并生成相应的故障根因检测模型。
在本实施例中，可以是根据专家知识经验结合各个样本单元中的模型样本，对每个样本单元添加根因标签，也可以是根据过往的处理经验结合样本单元中的模型样本的各项参数，对样本单元添加根因标签。
S14、对待诊断网络性能参数进行处理，得到待诊断样本，分别计算待诊断样本与每个模型样本的相似度值。
对待诊断的网络性能参数通过上述步骤中相同的处理方式得到待诊断样本，分别计算待诊断样本与样本单元中的每个模型样本的相似度值。
比如，计算待诊断样本与模型样本的距离度量，作为待诊断样本和模型样本的相似度值，在本实施例中，可以通过计算待诊断样本与模型样本的欧氏距离，作为待诊断样本和模型样本的相似度值，欧氏距离的值越小，相似度值越高。
S15、获取与待诊断样本相似度最高的模型样本对应的样本单元的根因标签，将根因标签作为待诊断样本的故障根因。
在本步骤中，将与待诊断样本相似度最高的模型样本的样本单元的根因标签作为待诊断样本的故障根因。
如图2所示，本发明实施例还提供了一种网络故障根因检测方法，与图1所示检测方法相比，区别在于，对每个样本单元添加根因标签，生成根因检测模型，包括：S21、根据样本单元中模型样本的数量对样本单元进行排序。
结合上述实施例，在本实施例中，根据样本单元中的模型样本的数量，对样本单元进行排序，模型样本数量越多的样本单元，排序越靠前。
S22、将排序高于预设阈值的样本单元设置为核心单元，采用基于密度的聚类算法将每个核心单元与其他样本单元进行合并，得到多个样本单元类簇。
在本实施例中，由于实际应用过程中很多相近的特征其根因相同，因此需要把一些相近的神经元进行合并以表示只是症状现象略有不同的同一类故障。将排序高于预设阈值的样本单元设置为核心单元，将其他样本单元分别与每个核心单元进行合并，使得具有相类似症状现象的样本单元组合成样本单元类簇。
S23、对每个样本单元类簇添加根因标签，生成根因检测模型。
通过对每个样本类簇添加根因标签，即可完成对该样本单元类簇下的所有样本单元的故障根因的确认，即对样本单元中的各个模型样本添加根因标签，提高对模型样本的故障根因的确定速率，并生成相应的故障根因检测模型。
在本实施例中，可以是根据专家知识经验结合各个样本单元类簇中的模型样本，对每个样本单元类簇添加根因标签，也可以是根据过往的处理经验结合样本单元中的模型样本的各项参数，对样本单元类簇添加根因标签。
如图3所示，本发明实施例提供了一种网络故障根因检测系统，网络故障根因检测系统包括处理器、存储器；处理器用于执行存储器中存储的故障根因检测程序，以实现第一方面中任一项实施例的故障根因检测方法。
对上述实施例中的系统或装置提供用于记录可以实现上述实施例的功能的软件程序的程序代码的存储介质，并通过系统或装置的计算机(或CPU或MPU)读取并执行存储在存储介质中的程序代码。
在这种情况下，从存储介质读出的程序代码本身执行上述实施例的功能，而存储程序代码的存储介质构成本发明实施例。
作为用于提供程序代码的存储介质，例如软盘、硬盘、光盘、磁光盘、CD-ROM、CD-R、磁带、非易失存储卡、ROM、以及类似物都可以使用。
上述实施例的功能不仅可以通过由计算机执行读出的程序代码来实现，而且也可以通过在计算机上运行的OS(操作系统)根据程序代码的指令执行的一些或全部的实际处理操作来实现。
此外，本发明实施例还包括这样一种情况，即在从存储介质读出的程序代码被写入被插入计算机的功能扩展卡之后，或者被写入和计算机相连的功能扩展单元内提供的存储器之后，在功能扩展卡或功能扩展单元中包括的CPU或类似物按照程序代码的命令执行部分处理或全部处理，从而实现上述实施例的功能。
本发明实施例提供了一种计算机可存储介质，计算机可存储介质存储有一个或者多个程序，一个或者多个程序可被一个或者多个处理器执行，以实现第一方面中任一项实施例的故障根因检测方法。
本发明的上述技术方案与现有技术相比具有如下优点：本发明实施例通过获取不同小区的网络性能参数，通过聚类将网络性能参数生成的模型样本映射到同一样本单元，对样本单元分别添加根因标签，生成根因检测模型，通过根因检测模型对待诊断网络性能参数进行诊断，实现对无线通信网络的故障根因的快速诊断，提高故障的诊断效率。
最后应说明的是：以上实施例仅用以说明本发明的技术方案，而非对其限制；尽管参照前述实施例对本发明进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或 者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本发明各实施例技术方案的精神和范围。

Claims (10)

  1. 一种网络故障根因检测方法，其中，所述检测方法包括：
    获取多个网络性能参数，分别对每个所述网络性能参数进行处理，得到模型样本；
    对所有所述模型样本进行细粒度聚类，将相似度大于预设阈值的模型样本映射到同一样本单元，得到多个样本单元；
    对每个样本单元添加根因标签，生成根因检测模型；
    对待诊断网络性能参数进行处理，得到待诊断样本，分别计算所述待诊断样本与每个所述模型样本的相似度值；
    获取与所述待诊断样本相似度最高的所述模型样本对应的样本单元的根因标签，将所述根因标签作为所述待诊断样本的故障根因。
  2. 根据权利要求1所述的网络故障根因检测方法，其中，所述对每个样本单元添加根因标签，生成根因检测模型，包括：
    根据样本单元中模型样本的数量对样本单元进行排序；
    将排序高于预设阈值的样本单元设置为核心单元，采用基于密度的聚类算法将每个核心单元与其他样本单元进行合并，得到多个样本单元类簇；
    对每个样本单元类簇添加根因标签，生成根因检测模型。
  3. 根据权利要求1所述的网络故障根因检测方法，其中，所述对所有所述模型样本进行细粒度聚类，将相似度大于预设阈值的模型样本映射到同一样本单元，包括：
    将所有所述模型样本输入自组织映射神经网络的输入层；
    通过自组织映射神经网络的竞争层完成对所有所述模型样本的细粒度聚类，将相似度大于预设阈值的模型样本映射到同一所述样本单元中。
  4. 根据权利要求1所述的网络故障根因检测方法，其中，所述分别对每个所述网络性能参数进行处理，得到模型样本，包括：
    通过主成分分析方法对每个所述网络性能参数进行降维处理，得到所述模型样本。
  5. 根据权利要求4所述的网络故障根因检测方法，其中，所述分别对每个所述网络性能参数进行降维处理之前，还包括：
    对所有所述网络性能参数进行过滤，剔除所述网络性能参数中的异常值。
  6. 根据权利要求4所述的网络故障根因检测方法，其中，所述检测方法还包括：
    将所述模型样本作为参考模型样本；
    对所有参考模型样本分别进行数据标准化处理，得到所述模型样本。
  7. 根据权利要求1所述的网络故障根因检测方法，其中，所述对待诊断网络性能参数进行处理，得到待诊断样本，包括：
    通过主成分分析方法对所述待诊断网络性能参数进行降维处理，得到所述待诊断样本。
  8. 根据权利要求1～7中任一所述的网络故障根因检测方法，其中，所述计算所述待诊断样本与每个所述模型样本的相似度值，包括：
    分别计算待诊断样本与每个所述模型样本的欧氏距离，作为所述相似度值。
  9. 一种网络故障根因检测系统，其中，所述网络故障根因检测系统包括处理器、存储器；所述处理器用于执行所述存储器中存储的故障根因检测程序，以实现权利要求1～8中任一项所述的故障根因检测方法。
  10. 一种计算机可存储介质，其中，所述计算机可存储介质存储有一个或者多个程序，所述一个或者多个程序可被一个或者多个处理器执行，以实现权利要求1～8中任一项所述的故障根因检测方法。
PCT/CN2019/111750 2018-11-26 2019-10-17 一种网络故障根因检测方法、系统及存储介质 WO2020108159A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811420214.9 2018-11-26
CN201811420214.9A CN111224805A (zh) 2018-11-26 2018-11-26 一种网络故障根因检测方法、系统及存储介质

Publications (1)

Publication Number Publication Date
WO2020108159A1 true WO2020108159A1 (zh) 2020-06-04

Family

ID=70827403

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/111750 WO2020108159A1 (zh) 2018-11-26 2019-10-17 一种网络故障根因检测方法、系统及存储介质

Country Status (2)

Country Link
CN (1) CN111224805A (zh)
WO (1) WO2020108159A1 (zh)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111404603B (zh) * 2020-06-03 2020-11-27 广东电网有限责任公司佛山供电局 一种基于最大匹配度的通信网故障快速定位方法
CN114389940A (zh) * 2020-10-20 2022-04-22 华为技术有限公司 故障恢复预案确定方法、装置及系统、计算机存储介质
CN114548195A (zh) * 2020-11-26 2022-05-27 中兴通讯股份有限公司 一种数据标注处理方法、装置、存储介质及电子装置
CN113079043B (zh) * 2021-03-25 2021-12-10 浙江大学 一种基于码本的网络故障诊断方法
WO2022266833A1 (zh) * 2021-06-22 2022-12-29 华为技术有限公司 一种根因分析方法及相关设备
CN113868008A (zh) * 2021-10-14 2021-12-31 中国建设银行股份有限公司 一种异常处理方法及装置
CN117033053A (zh) * 2023-08-14 2023-11-10 广州诺芯软件科技有限公司 应用于汽车营销管理系统软件的数据分析方法及系统

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106202922A (zh) * 2016-07-06 2016-12-07 吴本刚 一种基于聚类算法的变压器故障诊断系统
CN107222511A (zh) * 2017-07-25 2017-09-29 深信服科技股份有限公司 恶意软件的检测方法及装置、计算机装置及可读存储介质
CN107528823A (zh) * 2017-07-03 2017-12-29 中山大学 一种基于改进的K‑Means聚类算法的网络异常检测方法
US20180268296A1 (en) * 2016-06-02 2018-09-20 Tencent Technology (Shenzhen) Company Limited Machine learning-based network model building method and apparatus

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2014040633A1 (en) * 2012-09-14 2014-03-20 Huawei Technologies Co., Ltd. Identifying fault category patterns in a communication network
CN103914735B (zh) * 2014-04-17 2017-03-29 北京泰乐德信息技术有限公司 一种基于神经网络自学习的故障识别方法及系统
CN105487009A (zh) * 2015-11-19 2016-04-13 上海电机学院 基于k均值的RBF神经网络算法的电机故障诊断方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180268296A1 (en) * 2016-06-02 2018-09-20 Tencent Technology (Shenzhen) Company Limited Machine learning-based network model building method and apparatus
CN106202922A (zh) * 2016-07-06 2016-12-07 吴本刚 一种基于聚类算法的变压器故障诊断系统
CN107528823A (zh) * 2017-07-03 2017-12-29 中山大学 一种基于改进的K‑Means聚类算法的网络异常检测方法
CN107222511A (zh) * 2017-07-25 2017-09-29 深信服科技股份有限公司 恶意软件的检测方法及装置、计算机装置及可读存储介质

Also Published As

Publication number Publication date
CN111224805A (zh) 2020-06-02

Similar Documents

Publication Publication Date Title
WO2020108159A1 (zh) 一种网络故障根因检测方法、系统及存储介质
CN108415789B (zh) 面向大规模混合异构存储系统的节点故障预测系统及方法
CN111612038B (zh) 异常用户检测方法及装置、存储介质、电子设备
CN108052528A (zh) 一种存储设备时序分类预警方法
CN111612041B (zh) 异常用户识别方法及装置、存储介质、电子设备
Mao et al. Feature representation using deep autoencoder for lung nodule image classification
CN111553127A (zh) 一种多标记的文本类数据特征选择方法及装置
CN107391369A (zh) 一种基于数据筛选和数据过采样的跨项目缺陷预测方法
CN116662817B (zh) 物联网设备的资产识别方法及系统
CN109257383A (zh) 一种bgp异常检测方法及系统
Jinyin et al. Fast density clustering algorithm for numerical data and categorical data
CN113569920B (zh) 基于自动编码的第二近邻异常检测方法
CN112732690B (zh) 一种用于慢病检测及风险评估的稳定系统及方法
Xu et al. Hybrid label noise correction algorithm for medical auxiliary diagnosis
CN113657022B (zh) 一种芯片故障识别方法及相关设备
CN115017988A (zh) 一种用于状态异常诊断的竞争聚类方法
Lu et al. A modified active learning intelligent fault diagnosis method for rolling bearings with unbalanced samples
Liang et al. Disk Failure Prediction Based on SW-Disk Feature Engineering
CN112768090A (zh) 一种用于慢病检测及风险评估的过滤系统及方法
CN114528906A (zh) 一种旋转机械的故障诊断方法、装置、设备和介质
CN115438239A (zh) 一种自动化异常样本筛选的异常检测方法及装置
CN114077663A (zh) 应用日志的分析方法及装置
CN113934585A (zh) 磁盘故障预测方法及装置、设备、存储介质
CN112884167B (zh) 一种基于机器学习的多指标异常检测方法及其应用系统
Wang et al. Deep Unsupervised Workload Sequence Anomaly Detection with Fusion of Spatial and Temporal Features in the Cloud

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19889155

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 03/11/2021)

122 Ep: pct application non-entry in european phase

Ref document number: 19889155

Country of ref document: EP

Kind code of ref document: A1