jingtt/varclushi

A Python package for variable clustering

varclushi's Issues

AttributeError : 'VarClusHi' object has no attribute 'clusters'

Hi team,
How can i recify the above error when I try to get
Demo1_vc.info

Any idea?

DIfferent results on Each fresh run

I noticed that this algorithm gives different results, every time I restart my jupyter notebook kernel/start a new session.
This algorithm is not using any Random functions, what can be the reason behind this issue?

AttributeError: 'Rotator' object has no attribute 'rotate'

---

AttributeError Traceback (most recent call last)
in ()
----> 1 demo1_vc.varclus()

/usr/local/envs/py3env/lib/python3.5/site-packages/varclushi/varclushi.py in varclus(self, speedup)
213
214 if self.speedup is True:
--> 215 return self._varclusspu()
216
217 ClusInfo = collections.namedtuple('ClusInfo', ['clus','eigval1','eigval2','pc1','varprop'])

/usr/local/envs/py3env/lib/python3.5/site-packages/varclushi/varclushi.py in _varclusspu(self)
168 clus1, clus2 = [], []
169 rotator = Rotator()
--> 170 r_eigvecs = rotator.rotate(pd.DataFrame(c_eigvecs), 'quartimax')[0].values
171
172 comb_sigma1 = math.sqrt(np.dot(np.dot(r_eigvecs[:, 0], split_corrs.values), r_eigvecs[:, 0].T))

AttributeError: 'Rotator' object has no attribute 'rotate'\

After running the wine example, I am receiving the error: AttributeError: 'Rotator' object has no attribute 'rotate'

Thanks

Number of clusters formed is 0

Hi,

I am trying to incorporate this method as a feature selection process, But when I used this algorithm, I am only able to find 0 clusters (There are round 600 features and 800K train samples in my data). Is there any way this problem can be solved.

Thanks.

Does this work on Spark or modin.pandas or H2O?

Does this work on anything other than pandas? I've tried but gives different errors for each, will there be any chance of running it in a distributed environment?

Enhancement: Add a clustering version that is compartible with scikit-learn transformer

I have used the package and converted it to work like a scikit-learn transformer. Would you want to collaborate and see how to upgrade it to use scikit-learn in the back?

There are benefits to that. I did that to chain it with other transformers in my data reduction pipeline.