Assessment of machine learning performance with limited test data

A computer program product for mitigating machine learning performance digression due to insufficient test data availability, the computer program product comprising:

• one or more computer-readable storage media and program instructions stored on the one or more computer-readable storage media, the program instructions comprising:
• instructions to receive a set of data, wherein the set of data comprises a plurality of records, each record of the plurality of records comprises one or more features and associated feature values;
• instructions to parse the received set of data into a set of training data and a set of test data;
• instructions to generate a trained model by applying a model to the set of training data;
• instructions to apply the trained model to the set of test data;
• instructions to record a first set of performance values of the tested trained model, wherein the first set of performance values includes at least a test data accuracy percentage value;
• in response to the recorded test data accuracy percentage value of the trained model being above a threshold, instructions to associate the recorded test data accuracy percentage value with a performance baseline value;
• instructions to generate a set of modified test data, wherein instructions to generate the set of modified test data comprises applying a weight value to at least one of the associated feature values of the one or more features of each record of the plurality of records associated with the set of test data;
• instructions to apply the trained model to the set of modified test data;
• instructions to record a second set of performance values, wherein the second set of performance values includes at least a modified test data accuracy percentage value;
• instructions to associate the recorded modified test data accuracy percentage value with a performance results value;
• instructions to calculate a performance difference value based on the performance baseline value and performance results value;
• and
• instructions to generate a table for display, wherein the generated table includes at least:
• the applied weight value,
• a class distribution value, and
• the performance difference value.

A computer system for mitigating machine learning performance digression due to insufficient test data availability, the computer system comprising:

• one or more computer processors;
• one or more computer-readable storage media;
• program instructions stored on the computer-readable storage media for execution by at least one of the one or more processors, the program instructions comprising:
• instructions to receive a set of data, wherein the set of data comprises a plurality of records, each record of the plurality of records comprises one or more features and associated feature values;
• instructions to parse the received set of data into a set of training data and a set of test data;
• instructions to generate a trained model by applying a model to the set of training data;
• instructions to apply the trained model to the set of test data;
• instructions to record a first set of performance values of the tested trained model, wherein the first set of performance values includes at least a test data accuracy percentage value;
• in response to the recorded test data accuracy percentage value of the trained model being above a threshold, instructions to associate the recorded test data accuracy percentage value with a performance baseline value;
• instructions to generate a set of modified test data, wherein instructions to generate the set of modified test data comprises applying a weight value to at least one of the associated feature values of the one or more features of each record of the plurality of records associated with the set of test data;
• instructions to apply the trained model to the set of modified test data;
• instructions to record a second set of performance values, wherein the second set of performance values includes at least a modified test data accuracy percentage value;
• instructions to associate the recorded modified test data accuracy percentage value with a performance results value;
• instructions to calculate a performance difference value based on the performance baseline value and performance results value;
• and
• instructions to generate a table for display, wherein the generated table includes at least:
• the applied weight value,
• a class distribution value, and
• the performance difference value.