Uses of multivariate analysis methods

Abstract

The top quark study groups at the Tevatron accelerator use several analysis techniques to extract precision measurements of top quark properties and search for single top quark production. Cut-based analyses and advanced multivariate techniques are used to extract the signal from large backgrounds and to improve the sensitivity of measurements. As an example of the performance gain obtained with multivariate methods, the DØ single top quark search is presented. Likelihood discriminants, neural networks and decision trees have similar sensitivity, much improved over the cut-based analysis. Boosted decision trees may improve results even more.