predictive ability of matlab

I have original time series that experienced as some point in time abnormal values. I would like then to compare it to 12 other time series and learn which of these 12 most likely caused the abnormal values of original time series. I then need to store the "pattern" associated with abnormal values and if in the future that "pattern" starts appearing again, I will know what caused it. Are there techniques within Matlab that will allow me to do such thing?

[Greg Heath]

Unclear

 What and how long is the original?
 How many abnormal patterns and how long are they?
 Are the abnormal patterns similar?
 Why would any of the other 12 series be the cause of 
 abnormal values in the 1st?

Hope this helps

Greg

Answer 1

[John D'Errico]

This has absolutely nothing to do with the predictive ability of MATLAB. It is entirely to the ability of any code that you will write. Yes, it can be done in MATLAB. Virtually anything can. But no, I'm pretty sure that there is no explicit code that does exactly what you want. You need to write it.

I will say that a colleague of mine tried something like this many years ago, but with relatively little success.

You are essentially looking for a pattern recognition code. So I would do some reading in that area, were I you.

[Greg Heath]

The answers to the questions I asked would be necessary for formulating a pattern recognition code.

Greg