Feature Engineering

Endpoint: POST /feature-engineering

Request Parameters

File Upload

• file (required): The file to be processed. Supported formats include CSV, Excel, etc.

Form Parameters

• numeric_transforms (optional): A list of transformations to apply to numeric columns.
  • Supported values: scale, bin, poly.
• datetime_transforms (optional): A list of transformations to apply to datetime columns.
  • Supported values: year, month, day, hour.
• categorical_transforms (optional): A list of transformations to apply to categorical columns.
  • Supported values: onehot, freq.
• text_transforms (optional): A list of transformations to apply to text columns.
  • Supported values: tfidf, count.
• custom_transforms (optional): A list of custom transformations to apply. Each transformation is a dictionary with the following keys:
  • method: The transformation method (e.g., log, interaction, sqrt, etc.).
  • columns: The columns to apply the transformation to.
  • params: Additional parameters for the transformation.

Processing Logic

1. Read the uploaded file: The file is converted into a Pandas DataFrame.
2. Infer column types: The types of columns (numeric, datetime, categorical, text) are inferred automatically.
3. Apply numeric transformations (if specified):
   • scale: Standard scaling is applied to numeric columns.
   • bin: Numeric columns are binned into quantiles.
   • poly: Polynomial features are generated for numeric columns.
4. Apply datetime transformations (if specified):
   • year: Extracts the year from datetime columns.
   • month: Extracts the month from datetime columns.
   • day: Extracts the day from datetime columns.
   • hour: Extracts the hour from datetime columns.
5. Apply categorical transformations (if specified):
   • onehot: One-hot encoding is applied to categorical columns.
   • freq: Frequency encoding is applied to categorical columns.
6. Apply text transformations (if specified):
   • tfidf: TF-IDF vectorization is applied to text columns.
   • count: Count vectorization is applied to text columns.
7. Apply custom transformations (if specified): Custom transformations are applied based on the provided configuration.
8. Return the modified file: The transformed file is provided for download with the filename appended with _feature_engineering.

Example Request

curl -X POST "http://localhost:8000/feature-engineering" \
-F "file=@data.csv" \
-F "numeric_transforms=scale" \
-F "datetime_transforms=year" \
-F "categorical_transforms=onehot" \
-F "text_transforms=tfidf"