Method and system for selecting a learning model from among a plurality of learning models

A system for selecting a learning model defined by parameters and hyperparameters from among a plurality of learning models, said system comprising a plurality of learning models, hyperparameter values that are predefined before a learning process, parameter values that are calculated during said learning process, said system comprising a computing device including a model selection module and a model repository including a plurality of series of instruction each corresponding to a learning model of said plurality of learning models and each including hyperparameter values, said device being configured to:

• evaluate prediction performance of each learning model, by prediction performance indicators,
• evaluate a classification, by the model selection module, and provide, for each learning model, at least one classification value,
• select a said learning model, a prediction performance value and the classification value of which are greater than first predetermined threshold values,
• vary the hyperparameter values, from hyperparameter optimization indicators,
• select a said learning model when the prediction performance value and the classification value are greater than predetermined second threshold values and the hyperparameter value is greater than a predetermined threshold value.

A model selection module defined by parameters and hyperparameters, implemented by a computing device, said module comprising a model repository including a plurality of series of instructions each corresponding to a learning model and each including hyperparameter values that are predefined before a learning process, parameter values that are calculated during the learning process, said model selection module being configured to:

• evaluate prediction performance of each learning model, by prediction performance indicators,
• evaluate a classification, and provide, for each learning model, at least one classification value,
• select a said model, a prediction performance value and the classification value of which are higher than first predetermined threshold values,
• vary the hyperparameter values, from hyperparameter optimization indicators,
• select a said model when the prediction performance value and the classification value are higher than second predetermined threshold values and the hyperparameter value is higher than a predefined threshold value.