ordered sequence mining

[mavro]

Hello

I would like to work on my data, the data are ordered sequences.

I would like to classify and detect anomalies.

A technique like association rules is not the best option as I need a technique which takes into account the order of the sequences in order to detect the anomalies.
I need a technique that uses wild-cards/gaps in order to detect the probability of a sequence.

I saw some sequence mining techniques like sparse markov trannsducers but I can't find any software or open source to deal with the sequence mining problem.

If anyone knows a softawre to treat this problem I will be more than happy.

Thanks a lot

[phil123]

If you want to find something like association rules but that consider the time, you could check sequential rule mining. It is similar to association rules, but there is a sequential ordering and the input is a sequence database, and there can be some gap and some window size constraint. You could also check sequential pattern mining.

If you are interested by this, I have an open-source software offering several algorithms related to that:

http://www.philippe-fournier-viger.com/spmf/

You could check the example in the documentation section of the website to see if the input and output is what you are looking for.

However, i'm not sure that it would be enough for your problem

Philippe