2

I recently began learning about HMMs and wanted to ask about a possible application that would hopefully help me grasp the concepts.

One of the applications for a regular Markov Model is modelling the sequence of interactions for customers on a website, such as:

  1. Land on page
  2. Click image
  3. Add product to cart
  4. Bounce

So if we were to assume that these observations were created by certain 'types' of customers like 'bargain hunters' or 'people looking for inspiration', then we want to determine how many types exist and label these customers. Would that turn the problem into one suited for a HMM?

Tylerr
  • 1,532

2 Answers2

0

What you could consider is a HMM with an unknown number of hidden states or classes (the latter are the 'types' you are talking about). You might find a good review dealing with this specific topic in this article. And I found this question which is also in the topic. Then, once you have estimated the number of 'types', you can start classic HMM inference.

However, if you've just started learning to work with these probabilistic models, you should perhaps start by working with a fixed number of hidden states.

TheCG
  • 1,147
0

If you've just started, one of the best book I used for HMM is HMM for times series!

There is many applications of that model is the book also

Abdoul Haki
  • 803