Automatic selection of variables for a machine-learning model

A computer-implemented system for reducing an amount of computational resources consumed by a machine-learning model, the system comprising:

• a processing system;
• and
• computer-readable memory in communication with the processing system encoded with instructions for commanding the processing system to execute steps comprising:
• applying a machine-learning model to a first dataset to generate a first output, the first dataset comprising a randomized selection of variables,
• wherein the machine-learning model is trained prior to the applying;
• iteratively removing variables of the plurality randomized selection of variables from the first dataset resulting in a subset of data;
• for each iteration, (i) applying the machine-learning model to the subset of data, wherein the subset of data has one or more variables removed from the first dataset to generate a second output, and (ii) comparing the first and second outputs by determining whether a difference of the first and second outputs is less than a predetermined threshold value, resulting in iterative comparisons;
• determining a minimum set of variables that excludes variables having an impact below a predetermined threshold on the output of the machine-learning model based on the iterative comparisons,
• wherein applying the machine-learning model to the first dataset comprising only the minimum set of variables generates an output equal to the first output, wherein each of the output and the first output comprise at least one of a prediction, a decision, or a classification;
• applying, subsequent to the determining, the machine-learning model to new data to generate an output for the new data,
• wherein the new data includes the minimum set of variables;
• and
• periodically determining a new minimum set of variables based on repeated iterative comparisons with a subsequent randomized selection of variables.

A non-transitory computer-readable storage medium for reducing an amount of computational resources consumed by a machine-learning model, the computer-readable storage medium comprising computer executable instructions which, when executed, cause a processing system to execute steps including:

• applying a machine-learning model to a first dataset comprising medical data to generate a first output, the first dataset comprising a randomized selection of variables,
• wherein the machine-learning model is trained prior to the applying;
• iteratively removing variables of the randomized selection of variables from the first dataset resulting in a subset of data;
• for each iteration, (i) applying the machine-learning model to the subset of data, wherein the subset of data has one or more variables removed from the first dataset to generate a second output, and (ii) comparing the first and second outputs by determining whether a difference of the first and second outputs is less than a predetermined threshold value, resulting in iterative comparisons;
• determining a minimum set of variables that excludes variables having an impact below a predetermined threshold on the output of the machine-learning model based on the iterative comparisons,
• wherein applying the machine-learning model to the first dataset comprising only the minimum set of variables generates an output equal to the first output,
• wherein each of the output and the first output comprise at least one of a prediction, a decision or a classification;
• applying the machine-learning model to new data to generate an output for the new data,
• wherein the new data comprises medical data for a patient that includes the minimum set of variables and the output for the new data predicts whether the patient has a medical condition;
• and
• periodically determining a new minimum set of variables based on repeated iterative comparisons with a subsequent randomized selection of variables.