One CLass SVM is a well-known machine learning algorithm for novelty detection, but it is applicable only for a set of vectors not for time series data.
I read a paper recently of how to use (One-Class SVM) OCSVM with time series data by first converting the time series into a phase space then apply the projection into OCSVM [1]
Is there an implementation or library for such an algorithm to handle OCSVM with time series data?
[1] http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.653.2440&rep=rep1&type=pdf