NASDAQ prediction

In the first example you have experimented with predicting functions that can be expressed analytically. In this page you can use the demonstration applet to try to learn predicting financial data - in particular the NASDAQ stock index.

Again, this applet is only for illustration. Predicting next price is very hard; better success was reported when trying to predict supporting indicators, such as moving average in the simplest case, which can be also very useful for trading. Also note that predicting the direction of move is completely different task.

How to work with the applet

If you have not seen the first example, please explore it first - basic description is available there.
In this version, two possibilities are available:
- NASDAQ daily (i.e., end of day close values) for the whole year 2007 (from January 3 till December 31 of 2007) - 251 values
- NASDAQ weekly - data from July 5 of 2002 till January 4 of 2008 - 288 values
For your experiments it is useful to know that each of these time series have the following values: zero for interval below 0, close value in the interval 0-number of values, and again zero after the last known value. So for example to see the whole NASDAQ daily in 2007, set the interval from 0 to 251.
If you think that the prediction is very good, then try to predict value that is more distant. When trying to predict very close value, the network is usually approximating next (to be predicted) value by last input value and from large distance it may seem that the prediction is very exact.

Please wait until the applet is loaded.

Applet and description (c) Marek Obitko, 2008; the neural network in the applet uses Java classes BPNeuron and BPNet
from NeuralWebspace, (c) Tomáš Vehovský, 1998, that were modified for the purposes of this applet.