The methods for variables selections which allows foroptimal representation of objects group in the data are studied. Neural network based Boolean factor analysis is compared with two statistical methods - linear factor analysis and hierarchical clustering. It is shown that Boolean factor analysis is the best solution for binary data dimensionality reduction. Advantage of Boolean factor analysis is its nonlinearity.
Index Terms:
Factor analysis, Boolean factor analysis, Clustering, Neural networks, Binary Data, Dimensionality Reduction
Citation:
Dusan Husek, Alexander A. Frolov, Hana Rezankov?, V?clav Sn?el, Pavel Y. Polyakov, "Some Remarks on Binary Data Grouping," dexa, pp.559-565, 2008 19th International Conference on Database and Expert Systems Application, 2008