Hey All, I've trained a svm model using the e1071 svm() function in R.
When testing using my test set, I get fairly good results in my eyes. It doesn't have to be perfect and the average error is around 1.2 units. For these purposes, this is more than acceptable.
My problem is I'm trying to use this model to predict new Values, however every time I run it I only get one value in return. 6.77523.
This is beyond frustrating. I'll be upfront and admit that I'm sorta poking around in the dark here. As in I have little formal training in this, I have a strong Math background but not much statistics.
I'm thinking it's very possible that my simulated data is not being constructed properly. That perhaps some values in my simulated are not actually simulating the real world, and thus not giving accurate values. But What I can't understand is why am I getting this single value? It would be one thing if I was just getting unrealistic values, but my simulated data has variation, I've randomly generated everything I felt fits a firm distribution and use averages (that are usually multiplied against a randomly generated value). It's getting late local time so I can't post any data I'm using at the moment but I'm stumped.
Is it possible that one feature is exerting a disproportional influence and this is what is flattening out my predictions?
It would be one thing if this was predicting a growing line, as the values grow over time, but this is simply predicting 6.77523 for every entry and I don't know where to begin.
[link][7 comments]