MULTIVARIATE-ANALYSIS METHODS TO TAG B-QUARK EVENTS AT LEP SLC

Multivariate analyses are applied to tag Z --> bbBAR events at LEP/SLC. They are based on the specific b-event shape caused by the large b-quark mass. Discriminant analyses, classification trees and neural networks are presented and their performances are compared. It is shown that the neural network approach, due to its nonlinearity, copes best with the complexity of the problem. As an example for an application of the developed methods the measurement of GAMMA(Z --> bbBAR) is discussed. The usefulness of methods based on the global event shape is limited by the uncertainties introduced by the necessity of event simulation. As a solution we present a double tag method which can be applied to many tasks of LEP/SLC heavy flavour physics.