Apparatus and method with system error prediction
An apparatus includes a processor configured to execute instructions, and a memory storing the instructions, which when executed by the processor configure the processor to generate system error prediction data using an error prediction neural network provided with one of a plurality of log data sequences generated by pre-processing a plurality of log data pieces of component log data of a system. The system error prediction data comprises information of a plurality of system errors occurring at a plurality of respective timepoints.
1 . An apparatus comprising:
a processor configured to execute instructions, and a memory storing the instructions, which when executed by the processor configure the processor to:
generate a plurality of log data sequences, comprising:
generating a plurality of updated log data pieces by eliminating redundant log data pieces comprising repeated log data pieces and stopwords from a plurality of log data pieces; and
generating the plurality of log data sequences by grouping the updated log data pieces according to a predetermined number of log data pieces;
generate system error prediction data using an error prediction neural network provided with one log data sequence of the plurality of log data sequences generated; and
control a system based on the generated system error prediction data
wherein the system error prediction data comprises information of a plurality of system errors predicted to occur at a plurality of different timepoints where the error prediction neural network output the system error prediction data for the plurality of system errors from a single log data sequence so that prediction of multiple systems errors from the single log data sequence during utilization of the error prediction neural network is enabled.
2 . The apparatus of claim 1 , wherein the information of the plurality of system errors comprises at least one of a predicted timepoint of error occurrence, a type of system error, and a possibility of error occurrence.
3 . The apparatus of claim 1 , wherein the execution of the instructions configures the processor to train the error prediction neural network based on training data in which the plurality of system errors are labeled in the single log data sequence.
4 . The apparatus of claim 3 , wherein the plurality of system errors comprises different error lead times.
5 . The apparatus of claim 1 , wherein the execution of the instructions configures the processor to generate an error handling report based on the system error prediction data.
6 . The apparatus of claim 5 , wherein the generating of the error handling report comprises outputting an integrated error handling report when there is a mutually associative relationship between the plurality of system errors included in the system error prediction data.
7 . The apparatus of claim 5 , wherein the generating of the error handling report comprises outputting a plurality of error handling reports respectively corresponding to the plurality of system errors in response to the plurality of system errors included in the system error prediction data being independent.
8 . The apparatus of claim 5 , wherein the generating of the error handling report comprises outputting an error prevention report in response to a possibility of an occurrence of a plurality of mutually associated system errors included in the system error prediction data being less than a threshold value.
9 . The apparatus of claim 1 , wherein the error prediction neural network is trained by labeling a plurality of system errors with different error lead times in the plurality of log data sequences so the prediction of multiple systems errors from the single log data sequence during utilization of the error prediction neural network is enabled.
10 . The apparatus of claim 1 , wherein the error prediction neural network comprises a single multi-label deep neural network (DNN) configured to output the system errors prediction data for the plurality of system errors from the single log data sequence, thereby reducing computational load during training and utilization of the neural network.
11 . A processor-implemented method, comprising:
generating a plurality of updated log data pieces by eliminating redundant log data pieces from a plurality of log data pieces;
generating a plurality of log data sequences by grouping the plurality of updated log data pieces according to a predetermined number of log data pieces; and
generating a trained error prediction neural network through training by labeling a plurality of system errors with different error lead times in the plurality of log data sequences; and
wherein the trained error prediction neural network comprises a multi-label deep neural network (DNN) configured to output, from a single log data sequence, information of the plurality of system errors predicted to occur at a plurality of different timepoints, thereby prediction of multiple systems errors from the single log data sequence during utilization of the error prediction neural network is enabled.
12 . The method of claim 11 , wherein each of the labeled plurality of system errors comprises at least one of an error-occurrence timepoint and a type of system error.
13 . The method of claim 11 , wherein the eliminating of the redundant log data pieces comprises:
eliminating stopwords included in the plurality of log data pieces; and
eliminating repeating log data pieces from a plurality of log data pieces from which the stopwords are eliminated.
14 . The method of claim 11 , wherein the error prediction neural network is trained by labeling a plurality of system errors with different error lead times in the plurality of log data sequences so the prediction of multiple systems errors from the single log data sequence during utilization of the error prediction neural network is enabled.
15 . A computing system, comprising:
a processor configured to execute instructions, and
a memory storing the instructions, which when executed by the processor configure the processor to:
generate a plurality of log data sequences, comprising:
generating a plurality of updated log data pieces by eliminating redundant log data pieces comprising repeated log data pieces and stopwords from a plurality of log data pieces; and
generating the plurality of log data sequences by grouping the updated log data pieces according to a predetermined number of log data pieces; and
generate system error prediction data using an error prediction neural network provided one log data sequence of the plurality of log data sequences; and
control a system based on the generated system error prediction data,
wherein a size of the system error prediction data changes according to a size of a log data sequence so that a number of predictable system errors changes as the size of the log data sequence changes, thereby enabling scalable error prediction adapted to system monitoring requirements,, and
wherein the error prediction neural network configured to output, from the one log data sequence, the system error prediction data for a plurality of system errors occurring at a plurality of different timepoints, where prediction of multiple systems errors from a single log data sequence during utilization of the error prediction neural network is enabled,
wherein the number of errors included in the system error prediction data changes according to an increase and a decrease of the number of log data pieces included in a log data sequence.
16 . The system of claim 15 , wherein the system error prediction data comprises information of a plurality of system errors occurring at a plurality of respective timepoints.
17 . The system of claim 16 , wherein the information of each system error comprises at least one of a predicted timepoint of error occurrence, a type of system error, and a possibility of error occurrence.
18 . The system of claim 15 , wherein the generating of the plurality of log data pieces includes an elimination of redundant information in the plurality of log data pieces.
19 . The system of claim 15 , wherein the generating of the plurality of log data pieces includes a grouping performed with respect to some of the plurality of log data pieces.
20 . The system of claim 15 , wherein the error prediction neural network is trained by labeling a plurality of system errors with different error lead times in the plurality of log data sequences of varying sizes so the prediction of multiple systems errors from the single log data sequence during utilization of the error prediction neural network is enabled and the number of predictable system errors scales according to the size of the input log data sequence, thereby enabling scalable error prediction adapted to system monitoring requirements.