[Traminer-users] how to identify entities that exhibit selected sequences?

["Nicolas S. Müller"]

Dear Robin,

I'm not sure if I completely understood your question, but I'll try to 
answer.

It is possible with TraMineR to look for specific event subsequences 
with the 'seqefsub' function. The procedure is explained on page 108 of 
the User's guide 
(http://mephisto.unige.ch/pub/TraMineR/Doc/1.4/TraMineR-1.4-Users-Guide.pdf).

You can then use the 'seqeapplysub' function on your sequence set, using 
the output of 'seqefsub' and the option 'method=presence', to identify 
which of your event sequences contain the subsequence of interest.

Good luck with your project.

Nicolas S. Müller

On 06/29/2010 10:40 PM, Robin Way wrote:
> Greetings. I am a new R user and new TraMiner user.
>
> The project on which I am about to embark is, I believe, in the domain 
> of event sequences. My primary question is, when I perform the 
> analysis and select a set of event sequences of primary interest, can 
> the TraMiner package help me identify which of the sampled observation 
> units exhibited that sequence in the historical event data used to run 
> the algorithm? Furthermore, is there a capacity to produce a ruleset 
> describing the selected event sequences on a new set of event data for 
> either the same observation units at a different point in time, or for 
> a different set of observation units?
>
> For instance,  my project is concerned with the quality control and 
> error detection of a set of machines. Let's say I have 20 machines 
> (sampled from a larger pool of machines) and my event sequence data is 
> a set of status codes from each machine recorded multiple times per 
> day over the course of a calendar month (for instance, January 2010). 
> Once I arrive at a set of sequences that differentiate machine-days on 
> which there was a significant machine outage, I would like to be able 
> to apply the rules associated with outages to the same machines or 
> other machines outside the sample on different time periods. The 
> prospective benefit is to help avoid future outages by identifying 
> sequences that appear to be occurring in advance of conditions leading 
> to future outages (albeit in a non-parametric fashion).
>
> Is this possible with the current TraMiner release? Are there 
> alternative methods of achieving this objective? Thank you very much 
> in advance.
>