MLKDE: Estimação de Densidade por Kernel Baseada em Máximo-Verossimilhança

Abstract

In this paper, we propose to solve the problem of estimating the pro bability density function of a random variable using Kernel density estimation, which depends on the bandwidth value, wich is the method parameter. Variati ons in the value of this parameter can generate different shapes for the associa ted distributions. Therefore, we propose a method to determine the best value to be used for bandwidth for Kernel density estimation, considering data sets and using the maximum-likelihood function as a reference. The results show that the proposed method achieves suitable results for different scenarios of using probability distributions. We evaluate the performance of the proposed method using a metric that selects the largest absolute difference between the points of the true probability density function and that estimated by the methods.