The other semi-supervised pipelines have numeric labels, which the HDBSCAN primitive can work with. jm1 has labels that are True or False, which the HDBSCAN step can't hande. This can be resolved by encoding the labels prior to the clustering step, although this will be a bit more complicated due to the fact that the clusterer is a dataset -> dataframe primitive, and anything running before the clustering step will have to be wrapped in the dataset_map primitive.
The other semi-supervised pipelines have numeric labels, which the HDBSCAN primitive can work with.
jm1
has labels that areTrue
orFalse
, which the HDBSCAN step can't hande. This can be resolved by encoding the labels prior to the clustering step, although this will be a bit more complicated due to the fact that the clusterer is a dataset -> dataframe primitive, and anything running before the clustering step will have to be wrapped in thedataset_map
primitive.