Method and device for predicting errors in a computing system

A computing device for predicting errors in a computing system, comprising:

• a communication circuit configured to communicate with the computing system;
• a processor;
• and
• memory storing instructions configured to cause the computing device to:
• receive, by the communication circuit, log data generated by the computing system during operation of the computing system, the log data comprising tokens;
• input the tokens to a discriminator model which generates scores of the respect tokens, respectively, each score corresponding to a probability that a corresponding token is an anomaly token;
• perform a sharpening process on the scores, wherein the sharpening process adjusts the scores such that a standard deviation of the scores increases;
• determine an anomaly score based on the scores to which the sharpening process has been performed;
• and
• determine a likelihood of future occurrence of an error in the computing system based on the anomaly score,
• wherein the discriminator model is a neural network model configured to determine a score of the corresponding token.

A method, comprising:

• collecting log data of a computing system;
• generating training data comprising tokens by performing preprocessing on log data, among the collected log data, indicating normal state of the computing system;
• further generating the training data by performing a masking process that precludes some of the tokens from being identifiable;
• estimating replacement tokens for masked tokens using a generator model;
• inputting the training data, including the replacement tokens, to a discriminator model that performs inference thereon to:
• generate first prediction values of whether tokens of the training data, including the replacement tokens, are replacement tokens and
• generate second prediction values of which masking patterns were applied during the masking process;
• and
• updating parameters of the generator model and the discriminator model based on the first prediction values and the second prediction values.