This was a project I completed from the University of London to identify counterfeit notes using k-means using the data supplied in Banknote_authentication_dataset.csv. This file contains v1 and v2 values derived from preprocessed computer vision evaluation of the notes, but importantly does not contain any labels, all of which are stored in Banknote_original_data.csv.
The unlabeled data is first evaluated statistically, and then used to train the k-means clustering model using k=2. The results are compared with the original labelled data, giving a rather modest 65% accuracy. Not bad for only two variables.