Principal component analysis creates a series of orthogonal vectors within your feature space. The first PC is aligned to amount of the maximal amount of variance in the data, the second the second most, etc. PCA was used to reduce the feature space in my small data set.