I wish to find a way to calculate the transition probabilities in my Markov chain model.
Let's say a customer has three products [A B C] and in this example I wish to know the transition probability (up-sell) from B to C. I know that customers have different transition probabilities to move from State B to C based on some demographic variables.
One option would be using a decision tree model (like Random Forest) and I can say that (for example) the probability of moving from state B to C for male customers in New Yourk is 0.2.
One problem id that the some of transition probabilities (B to A, B to B and B to C) wouldn't be 1 and I have to normalize the data.
Would you advise a better approach for measuring this transition probabilities?