Thanks for the heads up. So - aside from trying them. Any opinion on said software? Did you find anything useful? Initially, the plan is to use this on tabular data and explore/mine that data further, but I'm open to see where it leads me. I may still go forward with a programmer as well as it's a small investment.
I'm sure they've been updated a bit since I tried them. Today, I would choose based upon the restrictions first--for testing-them-out purposes. If Rapidminer restricts the number of threads, I would choose knime if it doesn't restrict. At the time, I wanted parameter optimization. So I would choose whichever had the better optimizer, instead. (Grid search vs. Genetic Algo, for example.) I wanted a custom implementation of the kNN algorithm with parameters that I could optimize. It seemed easier to just code it myself rather than learn how to code it in one of the two mentioned apps. I think they are fine, subject to the above mentioned possible restrictions, for non-programmers to experiment/prototype with machine learning. They also have powerful (data analysis, for example) tools built in. Both have videos you can check out! I think you'll enjoy at least one of them! Keep us posted!
I used RM about 1-2 years ago. Don't remember the pricing, but I seem to remember it was a bit expensive. It is however VERY user friendly, at least in my opinion. Working with it, it is of course great for learning more about your data, but as has been mentioned: Beware of curve-fitting. Good luck!
Cool. Could I ask you how you used it and on what type of data? Did you learn anything new using it...? FWIW, I'm probably putting it on a hold for a little while due to current time constraints/priorities, but I will trial it and play around with it later this summer as there is a monthly trial which should give me an idea if it's worth paying for. Right now, I'm moving along on an ML project with a programmer, so I'm curious to see how that plays out.
I may have had bad luck, but this initial project was a disappointment and did not establish any predictive relationships/patterns in my data set. In fact, this algorithm did far, far worse than my own manual predictions on the same data set. I still think there's no substitute for market understanding, experience and creativity. Which is why I think most programmers will fail in conquering the market. Now, market understanding, experience, creativity AND programming - well, that's a different story.
Python. Recurrent neural network. I am not dismissing ML of course, but I'm stating that my initial project was a disappointment.