RE: Sequence clustering - order of the model

Tech-Archive recommends: Fix windows errors by optimizing your registry

Thanks a lot.
Now I understand the algorithm better, but I still have a couple of questions:
- how many sequence states are considered for "soft" clustering
- if the answer > 1 then: how come more that 1 state is considered if the used markov models are of 1st order?

I'm looking forward to your answer.
Scott

The algorithm is essentially a mixture of markov and cluster (EM) models. The input sequence is used to determine the cluster probability of the input across all clusters. For each cluster the input will have a probability from 0 to 100% to belong to that cluster. From the markov models contained in each cluster, weighted by the probability the input is contained in that cluster, the predicted state is derived. The same mechanism is true for the subsequent states, although I don't think that we recompute the cluster probabilities after each precdiction. We could (and arguably should) but the performance of such a scheme would be terrible.

Hello!

I've been reeding about sequence clustering algorithm and found that it uses 1st order Markov models to analyze sequences. What exactly does that mean? I know that 1st order models should only consider the last state in a sequence for prediction, but all the examples I've seen have more than one state in a sequence? How many states are therefore used for predictions?
And finally - does multiplication of transition probabilities for n previous transitions mean that n-order model is used or is this a different concept?
I would really appreciate your answers.
Scott
.