Social Network For Analytic Professionals
Clustering similar time series trends
Hello,
I am working on a Project on Time Series Data. I need to segment customer who have similar time series trends. Can someone guide me how to approach such Time Series Clustering issues.
I am working on a Project on Time Series Data. I need to segment customer who have similar time series trends. Can someone guide me how to approach such Time Series Clustering issues.
Replies to This Discussion
-
Permalink Reply by Ralph Winters on -
I would first start out by looking at the rolling 3, 6, and 12 month averages for the underlying variables, and then possibly turn it into a cluster analysis. Eventually, I would develop a formal time series model (ARIMA, regression, exponential etc.) on each group and then test for any parametric differences.
-Ralph Winters
-
Permalink Reply by Karteek on -
Ralph, Thanks for your advice. Actually I started doing 3 month MA and then clustering on these MA's.
We also tried doing a Log First Derivatives and Second Derivatives for 3 and 6 months and cluster on these derivatives.
Does SOMs help us more analyzing these time series data?
-
Permalink Reply by Ralph Winters on
-
SOM's is a neural network technique, and it is certainly worth a try. However, eventually you will be faced with having to explain the results which can sometimes be difficult when using this technique, especially if it is found to be at odds with the outcomes from other analyses.
-Ralph Winters
Featured
Advertisement
Analytic Jobs
Most Visited
© 2010 Created by Vincent Granville
