PredictorXL is one of the components in TraderXL Pro. Unfortunately due to licensing restrictions OLSOFT cannot distribute a time limited demo of PredictorXL. For the purpose of this review I will build a simple model to explore the features of this addin. The image below shows a screenshot of PredictorXL dialog box.
As you can see the options for PredictorXL are extremely straight forward. It is simply a matter of modifying the options for your desired application and then starting the training process. For the purposes of this review we are going to build a simple model to predict the share price for BHP.AX. This is by no means anywhere near a conclusive model and I am only using it to showcase the power of this software. The first thing I have done is to use BulkQuotesXL to download the closing share price and volume for BHP.AX. Secondly I downloaded the closing prices of the following futures contracts:
- Australian Dollar
- 1-Month Libor
- Aluminium
- Cooper
- Gold
- Silver
- Light Crude Oil
The screen shot below shows the data we will be using for this simple Neural Net model.
After you have collected your data it is simply a matter of clicking New in the NeuroXL predictor dialog box and then clicking learn. You can tweak the options as desired. The lag option allows you to define the number of values you wish to predict. I have found the best results are obtained when lag is set to one. If you click on the show process button you will be able to watch the learning process take place. I should note if you have a large data set and a low Error or % the learning process will take a significant amount of time. Below is a screenshot of NeuroXL Predictor learning about the input output pairs from our data set.
After learning is complete it is simply a matter of clicking on the predict button and choosing a desired cell for the prediction to be placed. The screenshot below shows what the actual value of BHP.AX was and also what our model predicted the value would be.
As you can see this is a fairly good result given such a small data set and the fact that the model was very simple. I look forward to exploring this program in more detail over the coming weeks.
Popularity: 37% [?]
