NOTE: In NeuroShell Trader, open the chart named “Example 14 – Prediction with paper trading and multiple issues” which is the basis for following example:

In this example we have done a number of things differently than the last example, starting with the fact that we have modeled 9 Select Sector SPDRs (ETFs) instead of GBP/USD. All SPDRs are loaded in the same chart but in different “chart pages”. You’ll note that you can switch from one sector to the next using the dropdown boxes at the top of the chart.

The neural network inputs are the same as the FOREX example, and each stock has it’s own neural net. However, here we have optimized the parameters of the inputs as well as the neural network thresholds. You can click on the plus (+) sign beside each rule to see that, although the rule itself is locked and must remain, the parameters inside the rules are not locked.

We have also used the paper trading feature as was the case with the trading strategy in Example 10. We held out several months for a paper trading period, thus instructing NeuroShell to keep the model that worked best on the paper trading period, even though the optimization goal was not to find the best paper trading performance. Using this paper trading feature, we feel more confident trading this model into the future than we otherwise would have.

Some other minor changes can be found on the Optimization tab. We instructed the optimizer to run exactly 3 minutes on each chart page (stock). We also used swarm optimization. Swarm optimization generally does better fine tuning when there are fewer parameters to optimize, compared to evolution strategies, which does better at finding tough solutions when there are many parameters.

Editors note: The GeneHunter optimization technique was used in previous versions of NeuroShell, but it was improved in release 6.0. However, we feel it has no particular advantage over evolution strategies and swarm.

Finally, we introduce the Chart Page Sum indicator, with which we built a combined equity curve for the whole portfolio of stocks in the chart. Notice that if you switch from one chart page to another, this indicator does not vary.