possible bug in adasyn.py

In Line 207 << x_class_gen.append([x_i + step * (X[x_i_nn[nn_z], :] - x_i) >>, seems to me X should be replaced with X_class as the Nearest Neighbor was re-computed based on X_class prior to entering the for loop.

I think that you are right. I will check on my own but you can submit a PR if you wish.

The bug was introduced when refactor SMOTE and ADASYN. Good catch.The visual plot make much more sense.