|
well. my question is more about temporal cluster. If I want to model the classification in a clustered way, namely, a sequence of +1 that is followed by sequence of 0 or sequence of -1, what is a practical technique I can apply? Thank you very much. I have attempted to include previous lags of attributes in SVM to model such behavior but the result is not good. |
|
Reading the post again, I need to clarify that 2 string of thoughts: 1. how to model temporal effect using SVM, I know there are a few techniques out there, which is one is the best if I make the assumption that the length of the temporal interdependence is loosely defined. 2. can clusting such as k-mean be applied to the original data before doing SVM? Thank you very much in advance. |
