What I’m asked to do is itself a bit vague, and it doesn’t help that my work isn’t experienced at understanding and using Fuzzy Logic. So I may not be too clear at saying my thoughts.
I got data on cars. Counting how many times they drove too fast etc. And I’m trying to score their driving. I have 5 InputVariables and 1 OutputVaraible. And I’m told to um… get the results distributed as a gaussian curve? Feed the input distributed as a gaussian cure? Something along those lines….
As a testing prototype, I currently have Input membership functions made from Ramps and Trapezoids. Using arbitary test numbers as percentage distribution, i.e. 25% of the the start data fits in “low”, 30% of the middle data is “medium”, 25% of the end data fits in “high”, and a bit in-between. Find the data points that fits this distribution, and draw the functions with those at the shape corners.
https://i.imgur.com/NpXc5HZ.png
…And then the output function is just simple placeholder shapes because we’re not sure what more we have to think about here.
https://i.imgur.com/ZFTBtDu.png
How would I improve this to get a gaussian distribution involved… somewhere? Does changing the functions to G-Curves improve anything as intended? Or is it more better to keep using the Trapezoids but with more fine distribution like 5 Terms?
Are the lazily made placeholders fine as an output function or they need to be more carefully calculated… somehow? Drawn with G-curves?
Does Gaussian Input (I just made that up) mean Gaussian Output after being processed by Fuzzy Logic?
Sorry for lots of questions here, not all of them are meant to be answered. Just random things that come to my mind. That’s not even all of them.
I just don’t know what’s the correct thing I should be focusing on, and thus the correct question to ask. So any advice that comes to your mind would be helpful.