DOC Fix FutureWarning in cluster/plot_kmeans_assumptions.html #24928
Conversation
| y_pred = KMeans(n_clusters=3, n_init="auto", random_state=random_state).fit_predict( | ||
| X_filtered | ||
| ) |
There was a problem hiding this comment.
| y_pred = KMeans(n_clusters=3, n_init="auto", random_state=random_state).fit_predict( | |
| X_filtered | |
| ) | |
| y_pred = KMeans(n_clusters=3, n_init=5, random_state=random_state).fit_predict( | |
| X_filtered | |
| ) |
There was a problem hiding this comment.
Looking at the results, we were not lucky with the initialization and we obtain a result that is not in line with the comment.
I propose to make 3 individual init and check that the result does not change compared to the version in main.
There was a problem hiding this comment.
I set n_init=10 for the two cases where the plot was different to the one in main. I left the other two with n_init="auto". Is that fine, or should I also set them to a specific value?
The plot now looks exactly like the one in main
|
I checked locally and we are good. I think that we should revisit this example because with pinging @ArturoAmorQ on this last point ;) since this example could be "tutorialized" in this regards. |
Reference Issues/PRs
Fix towards #24876 .
What does this implement/fix? Explain your changes.
Fixed future warning
The default value of n_init will change from 10 to auto in 1.4. Set the value of n_init explicitly to suppress the warningin plot_kmeans_assumptions.html.