Back to results
US 10,255,129 B2Granted
Fault diagnosis method and apparatus for big-data network system
AssigneeHuawei Technologies Co., Ltd.
InventorsXin Jiang, Hang Li
Life of this patent
Estimating the term…
Abstract
A fault diagnosis method for a big-data network system includes extracting fault information from historical data in the network system, to form training sample data, which is trained to obtain a deep sum product network model that can be used to perform fault diagnosis; and diagnosing a fault of the network system based on the deep sum product network model. The embodiments of the present application resolve a problem that it is difficult to diagnose a fault of a big-data network system.
Claims16
A fault diagnosis method for a network system, the method implemented by a computer of the network system, the computer comprising a computer processor, and the method comprising:
- obtaining, by the computer processor, historical data of the network system, the historical data being heterogeneous data, the heterogeneous data comprising structured data and non-structured data, the historical data comprising fault information, and the fault information describing a cause and a symptom of multiple faults of the network system;
- obtaining, by the computer processor, the fault information from a structured field of the structured data and data content of the non-structured data to determine multiple groups of values of fault-related random variables by discretizing information in the structured field according to a value range of the structured field in the structured data to determine the values of the fault-related random variables, one group of values of the fault-related random variables indicating an association relationship between a symptom and a cause of one fault of the network system, the fault-related random variables comprising a random variable of a first category and a random variable of a second category, the random variable of the first category representing a symptom of a fault of the network system, and the random variable of the second category representing a cause of the fault of the network system;
- using, by the computer processor, the multiple groups of values of the fault-related random variables as training sample data to train a deep sum product network model;
- assigning, the computer processor, a value to the random variable of the first category according to a symptom of a current fault of the network system;
- determining, by the computer processor, a marginal probability or a conditional probability of the random variable of the second category using the deep sum product network model and according to the assigned value of the random variable of the first category;
- and
- deducing, by the computer processor, a cause of the current fault according to the marginal probability or the conditional probability of the random variable of the second category.
4 more elements
A fault diagnosis method for a network system, the method implemented by a computer of the network system, the computer comprising a computer processor, and the method comprising:
- obtaining, by the computer processor, historical data of the network system, the historical data being heterogeneous data, the heterogeneous data comprising structured data and non-structured data, the historical data comprising fault information, and the fault information describing a cause and a symptom of multiple faults of the network system;
- obtaining, by the computer processor, the fault information from a structured field of the structured data and data content of the non-structured data to determine multiple groups of values of fault-related random variables, one group of values of the fault-related random variables indicating an association relationship between a symptom and a cause of one fault of the network system, the fault-related random variables comprising a random variable of a first category and a random variable of a second category, the random variable of the first category representing a symptom of a fault of the network system, and the random variable of the second category representing a cause of the fault of the network system;
- using, by the computer processor, the multiple groups of values of the fault-related random variables as training sample data to train a deep sum product network model by:
- generating a numerical matrix according to the multiple groups of values, each row of the numerical matrix corresponding to one fault of the multiple faults of the network system, and each column of the numerical matrix corresponding to one variable of the fault-related random variables;
- dividing the numerical matrix into m×n first submatrices of an equal size, both m and n being positive integers, and a sum of m and n being greater than or equal to two;
- obtaining m×n sum product network models in a distributed training manner and according to the m×n first submatrices;
- and
- determining the deep sum product network model according to the m×n sum product network models;
- assigning, by the computer processor, a value to the random variable of the first category according to a symptom of a current fault of the network system;
- determining, by the computer processor, a marginal probability or a conditional probability of the random variable of the second category using the deep sum product network model and according to the assigned value of the random variable of the first category;
- and
- deducing, by the computer processor, a cause of the current fault according to the marginal probability or the conditional probability of the random variable of the second category.
9 more elements
A fault diagnosis method for a network system, the method implemented by a computer of the network system, the computer comprising a computer processor, and the method comprising:
- obtaining, by the computer processor, historical data of the network system, the historical data being heterogeneous data, the heterogeneous data comprising structured data and non-structured data, the historical data comprising fault information, and the fault information describing a cause and a symptom of multiple faults of the network system;
- obtaining, by the computer processor, the fault information from a structured field of the structured data and data content of the non-structured data to determine multiple groups of values of fault-related random variables by extracting information from the data content of the non-structured data using at least one of a named-entity recognition algorithm, a keyword extraction algorithm, a relationship extraction algorithm, or a text categorization algorithm to determine the values of the fault-related random variables, one group of values of the fault-related random variables indicating an association relationship between a symptom and a cause of one fault of the network system, the fault-related random variables comprising a random variable of a first category and a random variable of a second category, the random variable of the first category representing a symptom of a fault of the network system, and the random variable of the second category representing a cause of the fault of the network system;
- using, by the computer processor, the multiple groups of values of the fault-related random variables as training sample data to train a deep sum product network model;
- assigning, by the computer processor, a value to the random variable of the first category according to a symptom of a current fault of the network system;
- determining, by the computer processor, a marginal probability or a conditional probability of the random variable of the second category using the deep sum product network model and according to the assigned value of the random variable of the first category;
- and
- deducting, the computer processor, a cause of the current fault according to the marginal probability or the conditional probability of the random variable of the second category.
4 more elements
A fault diagnosis method for a network system, the method implemented by a computer of the network system, the computer comprising a computer processor, and the method comprising:
- obtaining, by the computer processor, historical data of the network system, the historical data being heterogeneous data, the heterogeneous data comprising structured data and non-structured data, the historical data comprising fault information, and the fault information describing a cause and a symptom of multiple faults of the network system;
- obtaining, by the computer processor, the fault information from a structured field of the structured data and data content of the non-structured data to determine multiple groups of values of fault-related random variables by:
- discretizing information in the structured field according to a value range of the structured field in the structured data, to determine the values of the fault-related random variables;
- and
- extracting information from the data content of the non-structured data using at least one of a named-entity recognition algorithm, a keyword extraction algorithm, a relationship extraction algorithm, or a text categorization algorithm, to determine the values of the fault-related random variables, one group of values of the fault-related random variables indicating an association relationship between a symptom and a cause of one fault of the network system, the fault-related random variables comprising a random variable of a first category and a random variable of a second category, the random variable of the first category representing a symptom of a fault of the network system, and the random variable of the second category representing a cause of the fault of the network system;
- using, by the computer processor, the multiple groups of values of the fault-related random variables as training sample data to train a deep sum product network model;
- assigning, by the computer processor, a value to the random variable of the first category according to a symptom of a current fault of the network system;
- determining, by the computer processor, a marginal probability or a conditional probability of the random variable of the second category using the deep sum product network model and according to the assigned value of the random variable of the first category;
- and
- deducing, by the computer processor, a cause of the current fault according to the marginal probability or the conditional probability of the random variable of the second category.
7 more elements
A fault diagnosis apparatus for a network system, comprising:
- a non-transitory computer readable medium having instructions stored thereon;
- and
- a computer processor coupled to the non-transitory computer readable medium and configured to execute the instructions to:
- obtain historical data of the network system, the historical data being heterogeneous data, the heterogeneous data comprising structured data and non-structured data, and the historical data comprising fault information, which describes a cause and a symptom of multiple faults of the network system;
- obtain the fault information from a structured field of the structured data and data content of the non-structured data to determine multiple groups of values of fault-related random variables by discretizing information in the structured field according to a value range of the structured field in the structured data to determine the values of the fault-related random variables, one group of values of the fault-related random variables indicating an association relationship between a symptom and a cause of one fault of the network system, the fault-related random variables comprising a random variable of a first category and a random variable of a second category, the random variable of the first category representing a symptom of a fault of the network system, and the random variable of the second category representing a cause of the fault of the network system;
- use the multiple groups of values of the fault-related random variables as training sample data to train a deep sum product network model;
- assign a value to the random variable of the first category according to a symptom of a current fault of the network system;
- determine a marginal probability or a conditional probability of the random variable of the second category using the deep sum product network model and according to the assigned value of the random variable of the first category;
- and
- deduce a cause of the current fault according to the marginal probability or the conditional probability of the random variable of the second category.
7 more elements
A fault diagnosis apparatus for a network system, comprising:
- a non-transitory computer readable medium having instructions stored thereon;
- and
- a computer processor coupled to the non-transitory computer readable medium and configured to execute the instructions to:
- obtain historical data of the network system, the historical data being heterogeneous data, the heterogeneous data comprising structured data and non-structured data, and the historical data comprising fault information, which describes a cause and a symptom of multiple faults of the network system;
- obtain the fault information from a structured field of the structured data and data content of the non-structured data to determine multiple groups of values of fault-related random variables by discretizing information in the structured field according to a value range of the structured field in the structured data to determine the values of the fault-related random variables, one group of values of the fault-related random variables indicating an association relationship between a symptom and a cause of one fault of the network system, the fault-related random variables comprising a random variable of a first category and a random variable of a second category, the random variable of the first category representing a symptom of a fault of the network system, and the random variable of the second category representing a cause of the fault of the network system;
- use the multiple groups of values of the fault-related random variables as training sample data to train a deep sum product network model by:
- generating a numerical matrix according to the multiple groups of values, each row of the numerical matrix corresponding to one fault of the multiple faults of the network system, and each column of the numerical matrix corresponding to one variable of the fault-related random variables;
- dividing the numerical matrix into m×n first submatrices of an equal size, both m and n being positive integers, and a sum of m and n being greater than or equal to two;
- obtaining m×n sum product network models in a distributed training manner and according to the m×n first submatrices;
- and
- determining the deep sum product network model according to the m×n sum product network models;
- assign a value to the random variable of the first category according to a symptom of a current fault of the network system;
- determine a marginal probability or a conditional probability of the random variable of the second category using the deep sum product network model and according to the assigned value of the random variable of the first category;
- and
- deduce a cause of the current fault according to the marginal probability or the conditional probability of the random variable of the second category.
12 more elements
A fault diagnosis apparatus for a network system, comprising:
- a non-transitory computer readable medium having instructions stored thereon;
- and
- a computer processor coupled to the non-transitory computer readable medium and configured to execute the instructions to:
- obtain historical data of the network system, the historical data being heterogeneous data, the heterogeneous data comprising structured data and non-structured data, and the historical data comprising fault information, which describes a cause and a symptom of multiple faults of the network system;
- obtain the fault information from a structured field of the structured data and data content of the non-structured data to determine multiple groups of values of fault-related random variables by extracting information from the data content of the non-structured data using at least one of a named-entity recognition algorithm, a keyword extraction algorithm, a relationship extraction algorithm, or a text categorization algorithm to determine the values of the fault-related random variables, one group of values of the fault-related random variables indicating an association relationship between a symptom and a cause of one fault of the network system, the fault-related random variables comprising a random variable of a first category and a random variable of second category, the random variable of the first category representing a symptom of a fault of the network system, and the random variable of the second category representing a cause of the fault of the network system;
- use the multiple groups of values of the fault-related random variables as training sample data to train a deep sum product network model;
- assign a value to the random variable of the first category according to a symptom of a current fault of the network system;
- determine a marginal probability or a conditional probability of the random variable of the second category using the deep sum product network model and according to the assigned value of the random variable of the first category;
- and
- deduce a cause of the current fault according to the marginal probability or the conditional probability of the random variable of the second category.
7 more elements
A fault diagnosis apparatus for a network system, comprising:
- a non-transitory computer readable medium having instructions stored thereon;
- and
- a computer processor coupled to the non-transitory computer readable medium and configured to execute the instructions to:
- obtain historical data of the network system, the historical data being heterogeneous data, the heterogeneous data comprising structured data and non-structured data, and the historical data comprising fault information, which describes a cause and a symptom of multiple faults of the network system;
- obtain the fault information from a structured field of the structured data and data content of the non-structured data to determine multiple groups of values of fault-related random variables by:
- discretizing information in the structured field according to a value range of the structured field in the structured data to determine the values of the fault-related random variables;
- and
- extracting information from the data content of the non-structured data using at least one of a named-entity recognition algorithm, a keyword extraction algorithm, a relationship extraction algorithm, or a text categorization algorithm to determine the values of the fault-related random variables, one group of values of the fault-related random variables indicating an association relationship between a symptom and a cause of one fault of the network system, the fault-related random variables comprising a random variable of a first category and a random variable of a second category, the random variable of the first category representing symptom of a fault of the network system, and the random variable of the second category representing a cause of the fault of the network system;
- use the multiple groups of values of the fault-related random variables as training sample data to train a deep sum product network model;
- assign a value to the random variable of the first category according to a symptom of a current fault of the network system;
- determine a marginal probability or a conditional probability of the random variable of the second category using the deep sum product network model and according to the assigned value of the random variable of the first category;
- and
- deduce a cause of the current fault according to the marginal probability or the conditional probability of the random variable of the second category.
10 more elements
Dependencies are inferred from the claim text.