Linear Discriminant Analysis (LDA) is a classical statistical approach for supervised dimensionality reduction.
LDA takes consideration of all data objects (in costrast to only data objects in the boundaries of SVM) as well as the distribution of the classes. It projects high-dimensional data objects onto a low-dimensional data space where data objects are maximally class separable.
LDA computes an optimal projection by minimizing the within-class distance and maximizing the between-class distance simultaneously in order to achieve maximum class discrimination.