Cluster Analysis; Transaction Data; Binary Entropy; Sparse Matrices; Casino Player Tracking
Original Research Article
In this article, we propose an iterative procedure for clustering sparse high dimensional transaction datasets, specifically two casino player tracking datasets. a common problem in clustering sparse datasets with very large dimensions is that in addition to classical techniques of clustering being unable to provide useful results, latent variable methods used for clustering often do not lead to sufficient data reduction to yield useful and informative results either. initially, we propose a straightforward resorting of the full dataset and then define an information based sparsity index to subset the sorted data. this new dimension reduced dataset is less sparse, and thus, more likely to produce meaningful results using established techniques for clustering. Using this technique enables the clustering of two secondary datasets from two Las Vegas repeater market casino properties, which consist of the amount of money casino patrons gambled, termed coin-in, on a variety of slot machines.