Feature selection for machine learning classification problems: a recent overview
Sotiris B. Kotsiantis · 2011
A lot of candidate features are usually provided to a learning algorithm for pro- ducing a complete characterization of the classification task. However, it is often the case that majority of the candidate features are irrelevant or redundant to the learning task, which will deteriorate the performance of the employed learning algorithm and lead to the problem of overfitting. The learning accuracy and training speed may be significantly deteriorated by these superfluous features. So it is of fundamental importance to select the relevant and nec- essary features in the preprocessing step. This paper describes basic feature selection issues