|
Good afternoon, Please find the description of my problem: 1) I have 20 000 times series ( the length is 12 period). 2) I can create an Hidden Markov Chain for each of 10 000 time series in order to make simulation. 3) Around 50 % of the times series have a lot of zeros on it: I don't want to use the HMM on that because I don't have enough data for each, it can create a biais. However, I would like to create an average transition matrix based on the 10 000 time series. Do you have any idea of a methodology that I can implement for this problem ? 1) I can create a matrix 2*2 based on the two following state: a) I can create the states (zeros states and non zeros states) b) For each time series, I calculate the number of transitions. c) I can derive the average number of transition from the 10 000 times series. 2) Do you think I can do a methodology with different number of states ??? Thanks by advances Cheers Pep Gardiola ! |
