I would like to explain this question with an example. It is bit lengthy but please bear with me.

Let say, I have 3 events and each have a specific trend of time dependent data for a particular period. Using these 3 data or graphs or trends as training data, how can I classify a new trend of data online. This may fall in the category of pattern recognition or time series classification. According to my knowledge, Pattern recognition can be done only after collecting the complete data of that trend. But in my case I want it to classify within some initial data and not waiting till the end i.e. comparing the new row of data or trend with all the existing three events each time and within a particular window if it predominantly matches to any of the training trend then I am able to classify the event.

Please provide me some papers or documents which is relevant to my doubt.

Thanks in advance.