Computer program for performance testing of models

A non-transitory computer-readable medium including a computer program, wherein when the computer program is executed by one or more processors, the computer program allows the one or more processors to perform a method for testing performance of a model, the method comprising:

• generating a plurality of transformed data based on each of a plurality of data included in a data set, including:
• exchanging, based on a user input, a percentage of a first data of the plurality of data with the same first percentage of a second data of the plurality of data with each other at a selected degree to generate at least one of the plurality of transformed data;
• generating a test data set based on the plurality of data and the plurality of transformed data;
• testing the performance of the model at at least one time point of learning completion of the model, an operation of the model, or during learning of the model, by calculating the test data set by using the model,
• wherein the testing the performance of the model by calculating the test data set by using the model includes:
• testing learning performance of the model by calculating a first test data set by using the model at a predetermined epoch interval in response to a learning time point of the model,
• testing detection performance of the model by calculating the first test data set by using the model in response to a learning completion time point of the model, and
• performing testing for performance degradation of the model by calculating a second test data set by using the model in response to an operation time point of the model,
• wherein:
• the second test data set includes data additionally acquired in response to a second time point which is a time point after the first time point and a second test data subset constituted by transformed data for the data, and
• the second time point is a time point corresponding to the operation time point of the model, and
• wherein the performing the testing for the performance degradation of the model by calculating the second test data set by using the model in response to the operation time point of the model is performed based on whether the model determines that the transformed data included in the second test data subset is abnormal.

A computing device for testing performance of a model, comprising:

• a processor including one or more cores;
• a memory including program codes executable in the processor;
• and
• a network unit transceiving data with another computing device,
• wherein the processor is adapted to:
• generate a plurality of transformed data based on each of a plurality of data included in a data set, includes:
• exchange, based on a user input, a percentage of a first data of the plurality of data and the same percentage of a second data of the plurality of data with each other at a selected degree to generate at least one of the plurality of transformed data;
• generate a test data set based on the plurality of data and the plurality of transformed data;
• test the performance of the model at at least one time point of learning completion of the model, an operation of the model, or during learning of the model, by calculating the test data set by using the model;
• and
• generate the plurality of transformed data by processing the data set as an input of a pseudo anomaly generation model, including:
• calculating the first probability distribution and a first sample statistical amount for a first data set by using the first data set,
• calculating the second probability distribution and a second sample statistical amount for a second data set, and
• training the pseudo anomaly generation model,
• wherein the training the pseudo anomaly generation model includes:
• calculating an inter-distribution similarity between the first probability distribution and the second probability distribution, and
• determining whether to additionally perform the training of the pseudo anomaly generation model based on the inter-distribution similarity.