Dimensionality reduction techniques like PCA, tSNE among many are important unsupervised methods in helping reduce the features of our data by projecting the most important features such that the orginal inforamation is retained. In PCA this translates to maximising the variance on the projeced plane.