Also, I should mention, SVM is a great way to not only build functions,
but also visualise higher dimensions in 2D space:
http://scikit-learn.org/stable/modules/svm.html
(see 1.4.2. Regression)
Yes, overfitting is a concern. But that will happen with 2 or 8 or 80 features if you have only a limited qty of triggers (events). Decreasing features does does not decrease overfitting. Only increasing the qty of events improves this condition.
You can use cross-validation to compensate. This shuffles small dataset into larger, safer datasets:
http://scikit-learn.org/stable/modules/cross_validation.html Download this app: http://www.nutonian.com/products/eureqa/ |