iteration over a 0-d array in _nanrankdata

[nerdcha]

I gather that others have hit this (#12) but it still seems like a live issue, I'm afraid. It's hitting when X and y are ndarrays of what looks like the right shape.

There's a reproducible example on Iris data here.

[bittremieux]

I can confirm that this remains an issue. When this problem was previously reported (#8, #12) the issues appear to have been closed without any real resolution.

(In my case at least) the problem is caused by the inability to handle nan data as reported in #8. A (local) solution is to change the _nanrankdata method back to bottleneck.nanrankdata.
Commit 80a74c1 explicitly removed the dependency on bottleneck, but this isn't really a solution as the new functionality is broken.

[danielhomola]

@bittremieux's PR apparently fixed this. Let me know if the issue persist, but for the time being, I'll close this. Thanks again!

[Proteusiq]

Hello, the problem still remains. Unless I have missed the solution

boruta 0.1.5
Python 3.6 running on Docker

`TypeError                                 Traceback (most recent call last)
<ipython-input-124-be126db958b1> in <module>()
      1 # find all relevant features
----> 2 feat_selector.fit(X, y)

/opt/conda/lib/python3.6/site-packages/boruta/boruta_py.py in fit(self, X, y)
    199         """
    200 
--> 201         return self._fit(X, y)
    202 
    203     def transform(self, X, weak=False):

/opt/conda/lib/python3.6/site-packages/boruta/boruta_py.py in _fit(self, X, y)
    333         imp_history_rejected = imp_history[1:, not_selected] * -1
    334         # calculate ranks in each iteration, then median of ranks across feats
--> 335         iter_ranks = self._nanrankdata(imp_history_rejected, axis=1)
    336         rank_medians = np.nanmedian(iter_ranks, axis=0)
    337         ranks = self._nanrankdata(rank_medians, axis=0)

/opt/conda/lib/python3.6/site-packages/boruta/boruta_py.py in _nanrankdata(self, X, axis)
    500         Replaces bottleneck's nanrankdata with scipy and numpy alternative.
    501         """
--> 502         ranks = sp.stats.mstats.rankdata(X, axis=axis)
    503         ranks[np.isnan(X)] = np.nan
    504         return ranks

/opt/conda/lib/python3.6/site-packages/scipy/stats/mstats_basic.py in rankdata(data, axis, use_missing)
    264             return _rank1d(data, use_missing)
    265     else:
--> 266         return ma.apply_along_axis(_rank1d,axis,data,use_missing).view(ndarray)
    267 
    268 

/opt/conda/lib/python3.6/site-packages/numpy/ma/extras.py in apply_along_axis(func1d, axis, arr, *args, **kwargs)
    394     i.put(indlist, ind)
    395     j = i.copy()
--> 396     res = func1d(arr[tuple(i.tolist())], *args, **kwargs)
    397     #  if res is a number, then we have a smaller output array
    398     asscalar = np.isscalar(res)

/opt/conda/lib/python3.6/site-packages/scipy/stats/mstats_basic.py in _rank1d(data, use_missing)
    252 
    253         repeats = find_repeats(data.copy())
--> 254         for r in repeats[0]:
    255             condition = (data == r).filled(False)
    256             rk[condition] = rk[condition].mean()

TypeError: iteration over a 0-d array

`

[chansinhui]

Facing same issues here.

[FrancisHChen]

Same issues here:

Traceback (most recent call last):
boruta_selector.fit(dataX.values, dataY.values.ravel())
File "/spare/local/fchen/virtualenv/lib/python3.5/site-packages/boruta/boruta_py.py", line 201, in fit return self._fit(X, y)
File "/spare/local/fchen/virtualenv/lib/python3.5/site-packages/boruta/boruta_py.py", line 335, in _fit iter_ranks = self._nanrankdata(imp_history_rejected, axis=1)
File "/spare/local/fchen/virtualenv/lib/python3.5/site-packages/boruta/boruta_py.py", line 502, in _nanrankdata
ranks = sp.stats.mstats.rankdata(X, axis=axis)
File "/spare/local/fchen/virtualenv/lib/python3.5/site-packages/scipy/stats/mstats_basic.py", line 265, in rankdata
return ma.apply_along_axis(_rank1d,axis,data,use_missing).view(ndarray)
File "/spare/local/fchen/virtualenv/lib/python3.5/site-packages/numpy/ma/extras.py", line 395, in apply_along_axis
res = func1d(arr[tuple(i.tolist())], *args, **kwargs)
File "/spare/local/fchen/virtualenv/lib/python3.5/site-packages/scipy/stats/mstats_basic.py", line 253, in _rank1d
for r in repeats[0]:
TypeError: iteration over a 0-d array

[royalshan]

what is the solution of this issue? I am still facing the issue

[mejihero]

same issues:
TypeError Traceback (most recent call last)
in ()
1 feat_selector = BorutaPy(rf, n_estimators = 50, verbose = 2, random_state = 1)
2
----> 3 feat_selector.fit(X, y)

~\AppData\Local\Continuum\anaconda3\lib\site-packages\boruta\boruta_py.py in fit(self, X, y)
199 """
200
--> 201 return self._fit(X, y)
202
203 def transform(self, X, weak=False):

~\AppData\Local\Continuum\anaconda3\lib\site-packages\boruta\boruta_py.py in _fit(self, X, y)
333 imp_history_rejected = imp_history[1:, not_selected] * -1
334 # calculate ranks in each iteration, then median of ranks across feats
--> 335 iter_ranks = self._nanrankdata(imp_history_rejected, axis=1)
336 rank_medians = np.nanmedian(iter_ranks, axis=0)
337 ranks = self._nanrankdata(rank_medians, axis=0)

~\AppData\Local\Continuum\anaconda3\lib\site-packages\boruta\boruta_py.py in _nanrankdata(self, X, axis)
500 Replaces bottleneck's nanrankdata with scipy and numpy alternative.
501 """
--> 502 ranks = sp.stats.mstats.rankdata(X, axis=axis)
503 ranks[np.isnan(X)] = np.nan
504 return ranks

~\AppData\Local\Continuum\anaconda3\lib\site-packages\scipy\stats\mstats_basic.py in rankdata(data, axis, use_missing)
263 return _rank1d(data, use_missing)
264 else:
--> 265 return ma.apply_along_axis(_rank1d,axis,data,use_missing).view(ndarray)
266
267

~\AppData\Local\Continuum\anaconda3\lib\site-packages\numpy\ma\extras.py in apply_along_axis(func1d, axis, arr, *args, **kwargs)
392 i.put(indlist, ind)
393 j = i.copy()
--> 394 res = func1d(arr[tuple(i.tolist())], *args, **kwargs)
395 # if res is a number, then we have a smaller output array
396 asscalar = np.isscalar(res)

~\AppData\Local\Continuum\anaconda3\lib\site-packages\scipy\stats\mstats_basic.py in _rank1d(data, use_missing)
251
252 repeats = find_repeats(data.copy())
--> 253 for r in repeats[0]:
254 condition = (data == r).filled(False)
255 rk[condition] = rk[condition].mean()

TypeError: iteration over a 0-d array

[elnazsn1988]

Same error here - any solutions? it seems to be uncomfortable with the different Boruta versions