Neuro-Fuzzy Methods in Finance Applied to the German Stock Index DAX
Rudolf Kruse, Stefan Siekmann, Ralph Neuneier, Hans-Georg Zimmermann · Contributions to economics · 1998
We present an extended neuro-fuzzy system, combined with a semantic-preserving backpropagation-based learning algorithm, which makes effective use of prior knowledge and of historical data. After the network is initialized with a set of rules, the learning algorithm optimize the rule base without destroying the initial semantic. Due to a sparse initial network structure the effective number of parameters is small which prevents the network from overfitting. Our extended architecture enables us not only to change the shape and the position of the membership functions but also to apply typical pruning algorithms to delete or insert single premises. With insertion of premises, the rule base is improved by creation of new rules. Deletion of premises leads to a further reduction of the system complexity. Both strategies have a positive effect on the generalization performance. These methods are implemented in the software enviroment for neural networks SENN of Siemens Nixdorf Advanced Technologies. We have tested the neuro-fuzzy approach on two financial series. Here we present the encouraging results on the task to predict daily returns of the German Stock Index DAX.