How to extract class labels from pycaret setup?

Question

Using pycaret 2.3.10 for a multi-class problem.

from pycaret.classification import *
...
config = setup(X_null, 'YEAR', silent=True, fold=5, use_gpu=True, )

How can I extract the class labels? I know the classes, but I don't know the order that pycaret applies. In my case I have a number of years and I don't know whether class 1 corresponds to the first year, for examples. Or if I have 'Cats', 'Dogs', 'Monkeys', the classes are not ordinal. So, the ordering could depend on the dataset. Whatever appears first in the dataset. Or it is sorted?

Is there a way to extract the class labels in the correct order programmatically?

Answer 1

In PyCaret 2.X, you will get a list of information from the setup step so you can get Label Encoded using the below code

from pycaret.datasets import get_data
from pycaret.classification import setup
import pandas as pd
from pandas.io.formats.style import Styler

def get_conf(exp):
    for i in range(0, len(exp)):
        if isinstance(exp[i], list):
            if exp[i] and isinstance(exp[i][0], Styler):
                return exp[i][0].data

data = get_data('iris', verbose=False)
exp = setup(data = data, target = 'species', session_id=123, verbose=False)
conf_df = get_conf(exp)
target_mapping = conf_df[conf_df['Description'] == 'Label Encoded']['Value'].values
print(target_mapping)

In PyCaret 3.X, you will get Experiment object from the setup step so you can get Target Mapping using the below code

from pycaret.datasets import get_data
from pycaret.classification import setup
import pandas as pd

data = get_data('iris', verbose=False)
setup_conf = setup(data = data, target = 'species', session_id=123, verbose=False)
conf_df = pd.DataFrame(setup_conf._display_container[0])
target_mapping = conf_df[conf_df['Description'] == 'Target mapping']['Value'].values
print(target_mapping)