Peter Flom пишет:
Hello
In a job I have starting next week, the data will be a large set of
interrelated time series (I don't want to go into details because I am
not yet sure what is proprietary). There will be 19 time series for
each subject, all with a lot of points (thousands) and hundreds or
perhaps thousands of subjects. For each subject, the time series will
have multiple and probably some quite strong, relationships.
Any references to classification (clustering, tree-based methods,
discriminant analysis, functional data analysis or what-have-you) of
such long time series would be welcome, preferably without TOO much math
background (I had 3 semesters of calculus, but am much more interested
in applications than theorems and proofs)
you cen see site http://www.brodgar.com for the referencies on the DFA
(dynamic factor analysis) - it can calculate decomposition and scores
for each time series to be used in the classification
another way is to use some adaptive clustering (Kohonen's SOM and
especially ASOM are the choice)
in both cases there are ready software that can be used ... and, of
course, you can try R with a lot of packages on the issue :-))
Best regards,
Anatoly Saveleiv,
Kazan State Univ., Russia
----------------------------------------------
CLASS-L list.
Instructions: http://www.classification-society.org/csna/lists.html#class-l
