[FIXED] Error running Birch from scikit-learn in anaconda environment

Issue

I am applying scikit-learn's Birch clustering algorithm to the following DataFrame. Using spyder in an anaconda environment.

             column1        column2       column3           column4
0           -0.193389       0.064714      -0.244906         0.156697
1            1.679812      -0.106796       0.516472         6.651750
2            0.560031      -0.226183      -0.101638        11.535628
3            2.321758       0.107137       0.263732         1.163682
4            0.099532      -0.168835       0.087254         2.019619
              ...            ...            ...              ...
17919       -0.360455      -0.243193      -0.244906        -0.397145
17920       -0.357445      -0.237645      -0.244906        -0.397145
17921       -0.352748      -0.228988      -0.244906        -0.397145
17922       -0.353796      -0.230921      -0.244906        -0.346796
17923       -0.341864      -0.208929      -0.244906        -0.397145

The following code throws an error on the screen. I only have a problem with Birch, the rest of the algorithms I am using work correctly: AgglomerativeClustering, GaussianMixtures, Kmeans...

    from sklearn.cluster import Birch
    brc = Birch(n_clusters=4, threshold=0.15)
    cluster_predict = brc.fit_predict(df)
    print(cluster_predict)

ValueError: ndarray is not C-contiguous

Running exactly the same code in Google Colab, I get good results.

Solution

It seems to be a bug in scikit-learn version 1.1.1. I haven't looked for a solution. I simply changed to a lower version and it works.

Version 1.1.0 also has the same bug.

Answered By - Carola