On the Nonparametric Classification and Regression Methods for Multivariate EAS Data Analysis

Based on a special data analysis methodology developed for non-direct multivariate experiments, we present the expected accuracies of the KASCADE experiment on the elemental composition and primary energy estimation. The calculations were carried out with CORSIKA simulations using the ANI applied package routines. The detector responce also was simulated. The Neural Networks classification, the Bayesian Decision Making and the Nonparametric Regression approaches are used and compared.