Wf multiclass wine

Multiclass classification example using Octopus

This example demonstrates how to use Octopus to create a multiclass classification model. We will use the Wine dataset from sklearn for this purpose. The Wine dataset contains 3 classes (wine types) with 13 features. Please ensure your dataset is clean, with no missing values (NaN), and that all features are numeric.

import os

from octopus.example_data import load_wine_data
from octopus.modules import Tako
from octopus.study import OctoClassification
from octopus.types import FIComputeMethod, ModelName

### Load and Preprocess Data
df, features, targets = load_wine_data()

print("Dataset info:")
print(f"  Features: {len(features)} - {features}")
print(f"  Samples: {df.shape[0]}")
print(f"  Classes: {len(targets)} - {targets}")
print(f"  Target distribution: {df['target'].value_counts().sort_index().to_dict()}")

### Create and run OctoClassification for multiclass classification
# OctoClassification automatically detects multiclass (>2 classes) from the data
study = OctoClassification(
    study_name="multiclass_wine",
    studies_directory=os.environ.get("STUDIES_PATH", "./studies"),
    target_metric="AUCROC_MACRO",
    feature_cols=features,
    target_col="target",
    sample_id_col="index",
    stratification_col="target",
    outer_split_seed=1234,
    single_outer_split=0,  # only process first outer split, for quick testing
    workflow=[
        Tako(
            task_id=0,
            depends_on=None,
            description="step_1_tako_multiclass",
            n_inner_splits=5,
            models=[
                ModelName.ExtraTreesClassifier,
                ModelName.RandomForestClassifier,
                ModelName.XGBClassifier,
                ModelName.CatBoostClassifier,
            ],
            max_outliers=0,
            fi_methods=[FIComputeMethod.PERMUTATION],
            n_trials=20,
        ),
    ],
)

study.fit(data=df)

print("Workflow completed successfully!")
print(f"Results saved to: {study.output_path}")