I ran a test I Coolab (Single factor testing tool developed using the API - #26 by Whycliffes) of some nodes and have found around 200 of the “best” (see below), but is there a method to explore the correlation between the different nodes of the 200?
I would think that it is a value that in a ranking system there are not too many nodes that are too highly correlated, so I’m looking for a method to sort out the most correlated.
Hi, happy it worked. I have not ventured that far in finding a method to explore correlation between the different nodes. I have been using the “scoredfactor” and select the top200 factors and then create a ranking system with that. Through multiple iterations, I create a TOP10, TOP25, TOP50, TOP75, TOP100, etc. and run different similated strategies with each of the ranking. Its a long process, but has been working pretty well.
Let me know if you find a better way to select the combination of factor for the ranking with proper weight for each factors. For now, I have use equal weighting no matter how many factors I have. My best strategies have between 42 factors and 70 factors in the ranking system.
Steph
1 Like
I started learning about maching learning, to prep for the new module when its available (hopefully before christmas)! I beleive that this will be a lot better, but for now, the ranking systems is still the best way to look at sock from as many angle as possible.