Giters

eonu / sequentia

Scikit-Learn compatible HMM and DTW based sequence machine learning algorithms in Python.

Home Page:https://pypi.org/project/sequentia/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Ideas how to interpret classifications?

franz101 opened this issue · comments

commented

Is there a best practise for understanding which sub-sequences belong to which classes?
Or what factors the model made decide?

commented

It's fairly difficult to interpret the influence of sub-sequences on classification for both KNNClassifier and DeepGRU.

However with HMMClassifier + GMMHMM, you can inspect the state emission distributions and transition matrices for the HMMs belonging to each class. This will give you some idea about how each HMM is learning to model the sequences.

Note that HMMClassifier on v0.12.1/master is based on a version of hmmlearn that had major bugs in it. I suggest installing from the dev branch (which has been updated) until I make the next release.