1

I have a time series of position of a particle over time and I want to estimate model parameters of two HMM using this data (one for the x axis, the other for the y axis). I'm using the hmmlearn library, however, it is not clear to me how should I proced. In the tutorial, it states that this is the third way to use the library, however, when I use the code as bellow:

remodel = hmm.GaussianHMM(n_components=3, covariance_type="full", n_iter=100)
remodel.fit(X)  
Z2 = remodel.predict(X)

and X is the list of x-axis values, it returns

ValueError: Expected 2D array, got 1D array instead

What should I add to my data in order to turn it 2D?

augray
  • 3,043
  • 1
  • 17
  • 30
donut
  • 628
  • 2
  • 9
  • 23

2 Answers2

0

Caveat emptor: My understanding of HMM and this lib are based on a few minutes of googling and Wikipedia. That said:

To train an HMM model, you need a number of observes samples, each of which is a vector of features. For example, in the Wikipedia example of Alice predicting the weather at Bob's house based on what he did each day, Alice gets a number of samples (what Bob tells her each day), each of which has one feature (Bob's reported activity that day). It would be entirely possible for Bob to give Alice multiple features for a given day (what he did, and what his outfit was, for instance).

To learn/fit an HMM model, then, you should need a series of samples, each of which is a vector of features. This is why the fit function expects a two-dimensional input. From the docs, X is expected to be "array-like, shape (n_samples, n_features)". In your case, the position of the particle is the only feature, with each observation being a sample. So your input should be an array-like of shape n_samples, 1 (a single column). Right now, it's presumably of shape 1, n_samples (a single row, the default from doing something like np.array([1, 2, 3])). So just reshape:

remodel.fit(X.reshape(-1, 1))
augray
  • 3,043
  • 1
  • 17
  • 30
0

for me, the reshape method didn't work. I used the numpy's np.column_stack instead. I suggest you insert X = np.column_stack([X[:]]) before fitting the model, it should work out.

Ons Bouarada
  • 45
  • 7