Usage Guide

This page shows practical, copy-paste templates for using SmartML correctly.

All examples reflect the actual SmartEco.SmartML API and recommended workflows.

---

Installing SmartML

pip install SmartEco

Importing SmartML

SmartML is exposed through the SmartEco package.

from SmartEco.SmartML import load_dataset, run_training, SmartML_Inspect

Inspect Available Models

SmartML_Inspect()

This will show:

• Available classification models
• Available regression models
• Disabled models due to missing dependencies
• Unavailable models are automatically excluded during execution.

---

Using OpenML Datasets

SmartML can load datasets directly from OpenML.

X, y = load_dataset(
    openml_id=562,
    target="usr",
    subset=None,
)

print(f"Dataset loaded: X={X.shape}, y={y.shape}")

• OpenML datasets are treated the same as CSV datasets after loading.

---

Using Local CSV Datasets

SmartML supports standard CSV files.

X, y = load_dataset(
    csv_path="data/dataset.csv",
    target="label",
)

• CSV and OpenML datasets follow the same internal pipeline.

---

Subsampling Large Datasets

For large datasets, subsampling can be used to get quick baseline results.

X, y = load_dataset(
    csv_path="data/large_dataset.csv",
    target="label",
    subset=50000,
)

---

Selecting Specific Models

Run Only Selected Models

results = run_training(
    X_df=X,
    y_ser=y,
    task="classification",
    models=[
        "random_forest",
        "xgboost",
        "lightgbm",
        "smartknn",
    ],
)

---

Exclude Models

results = run_training(
    X_df=X,
    y_ser=y,
    task="regression",
    exclude=[
        "svr",
        "knn",
    ],
)

Model names are:

• Case-insensitive
• Normalized internally
• Task-specific

---

Sorting Results

SmartML does not auto-sort results.

• Classification

results.sort_values("macro_f1", ascending=False)

• Regression

results.sort_values("r2", ascending=False)

---

Full Code

from SmartEco.SmartML import load_dataset, run_training, SmartML_Inspect

SmartML_Inspect()


X, y = load_dataset(
    openml_id=562,
    target="usr",
    subset=None,
)

print(f"Dataset loaded: X={X.shape}, y={y.shape}")


results = run_training(
    X_df=X,
    y_ser=y,
    task="regression",
    exclude=[
        "tabnet",
        "nam",
    ],
    output_csv="results/benchmark.csv",
)

print(
    results.sort_values(
        "r2",
        ascending=False,
    )
)

---

Important Notes

• SmartML does not validate task–dataset correctness
• Wrong task selection will produce misleading results
• SmartML is not a production pipeline
• Benchmarks are dataset-specific