I have time series data. I am looking for a procedure to find if a particular pattern exists in the time series. To make it more clear, suppose I have a base time series in which the check for the input pattern is done. I give a part of some time series as input to the system and check the base time series if there is an approximate match for the input in the base time series.
I am performing this action in Knime. I cannot understand how HMM can fit into the problem. Is there any approach to achieve this task?
Best Answer
This sounds like a signal processing problem to me: e.g. Cross Correlation (CC), Coherence and similar approaches are frequently used in related problems, such as with radar technology. For example, using CC with a sliding window could be a solution to your problem (assuming that you know the duration of the reference sample you are searching for). Having said that, the exact algorithmic solution will depend on the type of series you are dealing with.
Sorry for posting this as an answer, but I'm still ranked too low for comments.