01-16-2014 09:11 AM
I have a fraud modeling project and my idea is to use monthly consumption (time series), first cluster the time series and then use the clusters as input in predictive model. The problem is that I have different observation windows, e.g. one time series from May 2011 to Sept 2012 and the other from December 2012 to June 2013, thus they differ from length and also timing. Can I take care of this in E. Miner? There is a sequence sliding option under TS similarity node, is this the one, or?
Thank you in advance!
01-21-2014 08:00 AM
The answer is Yes
The TS SIM node has the sequence sliding option. If you specify the value to “Index”, it will create the similarity (distance) measure between two series collected in two different time periods.