I am trying to implement HMM Learning/training with the continuous data set which is in sequential form. I have tried to construct a new HMM training with Gaussian Mixture/EM Algo, but I have been facing some issues, so i switch to hmmlearn library in python.
Overview about my data set
For e.g, A list contains normalized training data.
Total elements inside the list = 75
Each element inside the list is an np array.
The data inside the list is Vehicle driving data which is continuous.
'X_train = [[100*4], [100*4], [100*4], [100*4].........................[100*4]]'
len(X_train) = 75
X_Train[0] = [100*4]
X_Train.columns=['Speed','Angle', 'Acceleration1', 'Acceleration2']
Every [100*4] data received from vehicle at specific time intervals. Lets say X_train[0] is 15 to 30 seconds driving study data. X_train[1.] may be 15 to 30 seconds driving study data and so on .....
problem in hmmlearn
In hmmlearn library i would like to clarify some queries.
My objective is to get the parameters of the trained model for my dataset like initial prob, state transition prob, Gaussian weights, mean & covariance, so,.
Model created with 2 Hidden states with 2 Gaussian
model = hmm.GMMHMM(n_components=2, n_mix=2, covariance_type="diag", n_iter=100)
Question 1:
should I concatenate my overall data set into a single list as stated in the document Working with multiple seq from hmmlearn document?
If it is, then my training data set (X) will look like with size of 7500 *4, also, the lengths parameter will have a list like lengths = [100, 100, 100, ......100]
Is this the way fit should be implemented?
Question: 2
How can i check my EM Algo converges? I have seen the document for monitor_ function. But I am not able to understand.
Question: 3
After training, i would like to check the probability (using Forward Algorithm) of non-trained sequence of size 100*4. So what is the required function to apply forward algorithm using trained model parameters?
I have posted many question related to HMM training, But I have no satisfied answers based on my requirement. I hope hmmlearn library helps
