| CPC G06F 18/217 (2023.01) [G06F 18/285 (2023.01); G06N 20/20 (2019.01)] | 20 Claims |
|
1. A method comprising:
identifying, by a computing system, a set of one or more cross-validation parameters to be used for cross-validating a model to be used for generating a requested forecast, wherein the requested forecast includes a time series dataset and a forecast horizon identifying a number of time steps for which a forecast is to be made using the time series dataset;
identifying, by the computing system, an objective function to be minimized for determining optimal values for the set of one or more cross-validation parameters;
identifying, by the computing system, a set of constraints for one or more cross-validation parameters from the set of cross-validation parameters, wherein the objective function is represented as a set of penalty terms, and wherein:
a first penalty term in the set of penalty terms represents a cost of violation of a first constraint in the set of constraints on a first cross-validation parameter in the set of cross-validation parameters, wherein the first cross-validation parameter represents a left most fold cross-validation parameter for cross-validating the model; and
a second penalty term in the set of penalty terms represents a cost of violation of a second constraint in the set of constraints on a second cross-validation parameter in the set of cross-validation parameters, wherein the second cross-validation parameter represents a gap between the folds cross-validation parameter for cross-validating the model;
using, by the computing system, an optimization technique to determine the optimal values for the set of cross-validation parameters, wherein the optimal values for the set of cross-validation parameters is determined by:
determining one or more combinations of values to be assigned to the set of cross-validation parameters, wherein the set of cross validation parameters comprise the left most fold cross-validation parameter, the gap between the folds cross-validation parameter and a number of folds cross-validation parameter;
for each combination of values from the one or more combinations of values, computing a penalty value for the combination of values; and
determining the optimal values for the set of cross-validation parameters by selecting the combination of values from the one or more combinations of values that has the lowest penalty value; and
using, by the computing system, the optimal values determined for the set of cross-validation parameters to perform cross-validation of the model to be used for making the requested forecast.
|