Uma análise da aplicação de algoritmos de imputação de valores faltantes em bases de dados multirrótulo

Abstract

There are two types of databases that can be used by classification algorithms, which are multi and mono-label databases. In mono-label databases each example is associated with a single label, while in multi-label databases each example may be associated with more than one label simultaneously. The absence of values is a common problem in databases and to solve this problem there are imputation algorithms. Missing values imputation algorithms in multi-label databases are part of the preprocessing data stage so that multi-label classification algorithms can be applied. The treatment of incomplete data is made by imputation, where missing values are substituted by approximate values from other existing values in database. There are several algorithms that implement various ways to estimate values. In this context, the present study is an analysis of the aplication of missing values imputation algorithms in multi-label databases to check the efficiency of each on different databases with incomplete values. The algorithms Imputation by Mode, Mean, Median and Iterative KNN were tested, where the last one got the best results.