Time-series Data Mining

Poll

What types of time-series data mining you've done? [350 votes total] Anomaly detection (32) 9% Classification (49) 14% Clustering (50) 14% Forecasting (82) 23% Indexing (8) 2% Motif discovery (12) 3% Rule discovery (27) 8% Segmentation (31) 9% Summarization (19) 5% Other (13) 4% None so far (16) 5% None, but plan to do some soon (11) 3%

Comments

Joe Breeden, Multiple Time Dimensions
I find very little recognition in the academic literature that many data sets actually have more than one time dimension. I don't mean different time scales, but actual dimensions.

Almost all consumer behavior exists along two dimensions, (1) their responses to environmental effects, i.e. calendar date, and (2) the way their behavior changes with the length of time they have been a consumer of a given product. This latter dimension is called months-on-books in banking, but occurs in web site usage, insurance, online stores, health care, memberhsip programs, etc.

Is anyone aware of explicit treatment of this issue?

If you are, email to editor at kdnuggets.com