Recenlty, i have read some elegant jobs on nonparametric bayesian for time series analysis. In the attachment, i recommend one thourough work from NIPS, which is the top conference in ML. I believe it must be an interesting topic for the guys working in this field.