Підхід для покращення точності предиктивних моделей за рахунок інженерії ознак з оцінкою невизначеності

Abstract

The paper proposes an approach to improve the accuracy of bruise age prediction using uncertainty estimation. A Bayesian approach was applied to estimate the uncertainty of the predictions, and these estimates were integrated into the gradient boosting model as additional features. The proposed approach allowed us to improve the gradient boosting model by 8% compared to the model built solely on the original data. The work demonstrates that uncertainty estimation not only allows for a better understanding of the results, but also has the potential to improve the accuracy of predictive models.