PCA is a very common method for exploration and reduction of high-dimensional data. It works by making linear combinations of the variables that are orthogonal, and is thus a way to change basis to better see patterns in data. You either do spectral decomposition of the correlation matrix or singular value decomposition of the data