I have a streaming (not real-time) time series data comprising of thousand of multivariate objects (50 features each). Each object receives a new datastream of all features everyday. I want to track how similarity among objects changes over time. Could anyone share any research paper or throw some idea how to do this?