Uma análise sobre o impacto de dados faltantes no desempenho de métodos de aprendizado de máquina

Abstract

The occurrence of missing data is a recurrent problem and it has aroused the interest of researches over the last decades. Hence, many imputation methods have been proposed in recent years. In this dissertation, we present a study about the impact of the application of several imputation methods on the performance of machine learning algorithms, for both classification and regression. The result obtained shows that the imputation algorithms can have a relevant impact on the performance of classification and regression algorithms depending on the percentage of missing data. In addition, a model for the recommendation of data imputation algorithms is presented, which compares three classifiers (Random Forests, Gradient Boosting and Support Vector Machine) in the development of this task where both have good results.