I am trying to create a simple Neuro-Fuzzy system to predict race times (the data i'm using is form the 2010 motoGP series) based off of 5 inputs: Rider, No. of starting Riders, Track Length(can be used to identify individual tracks), Grid Position and Manufacturer.
The output should be a value in seconds that the rider would be expected to complete the track in.
However i am getting the error 'Xout must be finite' when using the GenFIS command. The data I have ranges from roughly 1700 seconds, to 3000 seconds for the expected output, I have 16 races to train the system and then 2 to test (very small dataset but the best i currently have access to) would I have to set a hard limit on the output and if so how would i go about doing that?
(I know this would be a crude workaround as its entirely possible a future race may take longer but it'll do for now)
I agree, you should consider a simpler example. So you can understand the complete architecture
Thanks for the help, the solution was extremely simple and turns out I was just being an idiot. In the data i have where one rider finished significantly behind the rest it was displayed as laps instead of a time (eg +3 laps), my system was not very happy with this input and was considering it infinite (hence x out must be finite error). A simple reminder for myself and others, check your data before spending hours attempting different methods.
- Badges
- Science topic