ScalerCrafter

The ScalerCrafter normalizes numerical features using various scaling techniques.

Overview

from mlfcrafter import ScalerCrafter

crafter = ScalerCrafter(
    scaler_type="minmax",
    columns=None  # Auto-select numerical columns
)

Parameters

scaler_type (str)

Default: "minmax"

Available Types: - "minmax": MinMaxScaler - scales features to [0,1] range - "standard": StandardScaler - standardizes features (mean=0, std=1) - "robust": RobustScaler - uses median and IQR, robust to outliers

columns (Optional[List[str]])

Default: None (auto-select all numerical columns)

Specific columns to scale. If None, automatically selects all numerical columns except target column.

Context Input

• data: Dataset to scale (required)
• target_column: Target column to exclude from scaling (optional)

Context Output

• data: Dataset with scaled numerical features
• scaler: Fitted scaler object for future use
• scaled_columns: Names of columns that were scaled
• scaler_type: Type of scaler used

Example Usage

from mlfcrafter import MLFChain
from mlfcrafter.crafters import *

# Scale all numerical columns with MinMax
pipeline = MLFChain()
pipeline.add_crafter(DataIngestCrafter("data.csv"))
pipeline.add_crafter(ScalerCrafter(scaler_type="minmax"))
pipeline.add_crafter(ModelCrafter())
result = pipeline.run()

# Standard scaling for specific columns
pipeline = MLFChain()
pipeline.add_crafter(DataIngestCrafter("data.csv"))
pipeline.add_crafter(ScalerCrafter(
    scaler_type="standard",
    columns=["feature1", "feature2"]
))
result = pipeline.run()

# Robust scaling (good for data with outliers)
pipeline = MLFChain()
pipeline.add_crafter(DataIngestCrafter("data.csv"))
pipeline.add_crafter(ScalerCrafter(scaler_type="robust"))
result = pipeline.run()