Which technique is most appropriate for reducing the dimensionality of a dataset while preserving the maximum variance?