Statistical sample sequence classification method for time series data e.g. stock market

摘要

The invention relates to a method of classifying a first series of statistical values having a given number of sample values, especially those of an electrical signal, by computer. A set of statistical values transmitted by a measuring signal of a dynamic system i.e. current share prices on the stock market, is modelled according to its probability density in order to provide a prediction of future values. A non-linear Markov process of the order m is suited to describe conditional probability densities. A neuronal network is trained in compliance with the probabilities of the Markov process according to the maximum likelihood principle, which is a learning rule in order to maximize the product of probabilities. For a predetermined number of values m arising from the past of the signal which is to be predicted, the neuronal network predicts a value in the future. Several steps in the future can be predicted by iteration. The order m of the non-linear Markov process acts as a parameter for improving the likelihood of the prediction. The order m corresponds to the number of past values which are important during modelling of conditional probability densities.