Financial forecasting of the German Stock Index (DAX)

Application Summary

Application category: Financial forecasting of the German stock index (DAX)

Keywords: time series forecasting, investment, neural networks

Company:

nodes GmbH
Warngauerstr. 31, 81539 Muenchen, Germany

Contact Email:

Rudolf_Suchar@nodes.de

Application Summary:

§ 1: Problem Description

We solved the daily and weekly forecasting for the DAX index.

Our plan is to use the product in our company for own investments and also to sell it to the other investment companies.

§ 2: Investigator's Background

The company has three major researchers on the field of neural networks, one of them having already the PhD on neural networks and other two finishing their PhD on the field of neural networks and related fields.

§ 3: How NNs were applied

Topology(s) used:

a) multilayer perceptron networks

b) modular multilayer perceptron networks

c) Elman networks

d) TLRN networks

Data description:

a) daily DAX index;

b) weekly DAX index;

c) related financial and economic time series;

d) technical indicators of the DAX index;

Data acquisition;

a) Datastream for daily and monthly time series;

b) VWD for real time data;

Data preprocessing/postprocessing

a) normalisation;

b) logarithm;

c) trend extraction;

d) linear and nonlinear supperposition of input series;

e) nonlinear methods for input series selection.

Training time required

a) ½....1 hours on Pentium100 PC's depending on the topology. The operating system is WindowsNT.

§ 4: Results that were obtained

We used as results estimator the directional test DT- the ratio between the number of predictions that correctly caught the direction of movement (increase/decrease) and the number of total predictions.

The best DT results are as follows:

70% for daily DAX predictions;

80% for weekly DAX predictions.

§ 5: How were the results applied

The results of the predictions are embedded in a global investment software which performs a stock market simulation.

Download this Application Summary in Word97 format.
If you don't have Word97....Download the Word Viewer.