Validation in self-organising data mining

Causality cannot be inferred from data analysis alone. Therefore, validation of input/output models generated by self-organising data mining means to decide if the pattern derived from data does exist, actually, or if it is only a stochastic one. By means of Monte Carlo Simulation a noise sensitivity characteristic for a given algorithm is identified. This characteristic provides external information required to validate any generated input/output model on the fly if it just models noise or not. If a model is stated not valid, there are some alternatives to improve its quality.