The Mahalanobis Distance for Feature Selection Using Genetic Algorithms: An Application to BCI

High dimensionality is a big problem that has been receiving a lot of interest from data scientists. Classification algorithms usually have trouble handling high dimensional data, and Support Vector Machine is not an exception. Trying to reduce the dimensionality of data selecting a subset of the original features is a solution to this problem. Many proposals have been applied and obtained positive results, including the use of Genetic Algorithms that has been proven to be an effective strategy. In this paper, a new method using Mahalanobis distance as a fitness function is introduced. The performance of the proposed method is investigated and compared with the state-of-the-art methods.