I am looking for a comparison of data-characteristics and classification methods, in particular for Support Vector Machines. As we know, non-linear separable data works well with SVMs. I would like to focus on time series data and one-class SVMs and the characteristics of the data.

What are characteristics of (time series) data on which one-class SVMs work very well, but other methods (and which methods) generally have problems?

For instance: time series data with a sinusoid shape which differs in frequency only (without extracting the frequency as a feature first) are hard to classify for SVMs. Change in mean and variance (again, without explicit feature extraction) are easy to detect.

Help and suggestions are very much appreciated!