Hidden Markov model example

hmm.py

from __future__ import division
import numpy
from sklearn import hmm

"""
X is a numpy array with shape (#samples, #sensors).
X = [[sensor1(0), sensor2(0)], [sensor1(1), sensor2(1)], ..., [sensor1(T), sensor2(T)]]

nc = number of latent states (3 would imply three strata)

"""

def Fit(X, nc):
	NUM_COMPONENTS = nc
	model = hmm.GaussianHMM(
			NUM_COMPONENTS, covariance_type="full", n_iter=1000)
	model.fit([X])
	hidden_states = model.predict(X)
	h = numpy.array(hidden_states)
	diff = h[1:] - h[:-1]
	return list(numpy.transpose(numpy.argwhere(diff != 0))[0])