I'm a fairly new RapidMiner user, so many thanks in advance for your assistance. I'm trying to set up an example for our team whereby we take windowed data and create a 'higher' or 'lower' prediction for forward periods. So for example, if the last 5 day's temperatures have been 28,29,34,29,30 degrees, we would like to know if the target period is likely to be higher or lower than the last period of the data (30 in this case), and if possible with what probability for an accurate target. So, to reiterate, we are not trying to estimate a future point temperature, just simply if it'll be higher or lower than what we've recently seen.
Thus far, I've successfully set up a simple linear regress model, but where my knowledge stops is how to qualify the forecasting performance according to the higher or lower target value. Could someone indicate a method for this?
Here are photos of what I have thus far.
Solved! Go to Solution.
The short answer is yes, you can easily predict the up or down trends but getting there will likely require Optmization and maybe a SVM with an RBF Kernel.
Go to my website and download the sample process at the end of the tutorial here: http://www.neuralmarkettrends.com/building-an-ai-financial-market-model-lesson-iv/
Then add your data and play around with it.