System and method for contextual event sequence analysis
AssigneeNEC Laboratories America, Inc.
InventorsJianwu Xu, Hui Zhang, Haifeng Chen, Tanay Kumar Saha
Life of this patent
Estimating the term…
Abstract
Systems and methods for contextual event sequence analysis of system failure that analyzes heterogeneous system event record logs are disclosed. The disclosure relates to analyzing event sequences for system failure in ICT and other computerized systems and determining their causes and propagation chains.
Claims20
A computer system for contextual event sequence analysis of system operation, comprising:
- a processor device;
- and
- a memory device operatively coupled to the processor device, the memory device configured to serve as an event record log source including digital event record logs, the processor device configured to execute program code stored on the memory device to:
- extract patterns from event records using an event pattern extractor;
- generate an event corpus from the extracted event patterns using an event identifier corpus generator;
- execute a machine learning algorithm to learn encoded latent representations of sequences of event types using a vector encoder coupled to the corpus generator;
- generate clusters of events using an event type sequence clusterer;
- generate automaton models of events ordered based on a context of events obtained from retrieved event records for the clusters using an automaton generator;
- and
- identify issues in the operation of a system using the automaton models to resolve the issues by locating the source of each issue and preventing the propagation of the issue through the system.
A computer program product comprising a non-transitory computer readable medium including instructions executable by a computer system instructing the system to:
- identify patterns within event record logs wherein a pattern represents a type of event;
- generate a corpus of event types;
- encode latent representations of event type sequences;
- learn latent representations of event type sequences;
- generate clusters of failure event types;
- generate an automaton model based on a context of events obtained from retrieved event records for each failure cluster;
- and
- order a plurality of automaton models.
A method for contextual event sequence analysis of system operation in a computerized system, comprising:
- identifying patterns within event record logs wherein a pattern represents a type of event;
- generating a corpus of event types;
- encoding latent representations of event type sequences;
- learning latent representations of event type sequences;
- generating clusters of event types;
- generating an automaton model for each cluster;
- and
- identifying issues in the operation of a system using the automaton models to resolve the issues by locating the source of each issue and preventing the propagation of the issue through the system.
Dependencies are inferred from the claim text.