Análise de Explicabilidade de um Modelo de Aprendizado de Máquina para Aplicações Industriais

Abstract

Machine learning models are used in numerous applications and, for some of them, it is desirable to be able to explain the output of the models. The search for model explainability lead to the development of Tree SHAP : a framework for tree-based models that calculates the contribution of each input feature to the predictions. In this paper, we train a regression model that predicts the yield value of seamless steel tubes at the end of the heat treatment process. We then perform a model explainability analysis to highlight the gains of such analysis in industrial applications.