No. 50 (00161) Family name : Zimmermann Given name : Jens Affiliation : Forschungszentrum Juelich, MPI Physik Muenchen Abbreviation : FZJ, MPI MUC E-mail address : zimmerm@mppmu.mpg.de Title : Statistical Learning Methods Authors : Jens Zimmermann Abstract : The statistical learning methods which are currently employed in the analysis of many experiments especially in the sector of high energy and astro particle physics can be categorized into three groups: - Decision Trees: for example CART or C4.5 - Local Density Estimators: for example k nearest neighbour search - Methods Based on Linear Discrimination: for example neural networks or support vector machines. Furthermore Meta-Learning strategies like Bagging (for example in Random Forests) or Boosting (for example in AdaBoost) have received much attention. The aim of this talk is to give a short overview over the different groups of statistical learning methods: How they work and how a classification or regression (the two applications of statistical learning methods: Separating two or more event classes from each other or estimating an unknown parameter from the observables) is finally done. Some Algorithms will be discussed in detail and their pros and cons will be studied. Finally some examples of successful application of statistical learning methods in running and for future experiments will be shown and the comparison of the algorithms among themselves and with parametric models - which are often the first choice - is discussed.