Feature Engineering Help

Create derived features — ratios, sums, products, means, log1p, polynomial interactions, longitudinal summaries, one-hot columns, or a wide→long reshape — saved as a new dataset for downstream analysis.

Requires: Numeric columns (plus a group column for one-hot).

Form fields accepted by this method's POST request

analysis_type=feature_engineering, dataset_id, fe_columns, fe_degree, fe_id_cols, fe_interaction_only, fe_name, fe_operation, fe_prefix

No dataset yet? Upload a file.
Parameters & Options
Ratio/difference use exactly two columns; sum/product/mean/longitudinal use two or more; log1p and one-hot apply per column; wide→long melts the selected timepoint columns.
The output is saved as a new derived dataset containing your original columns plus the engineered ones, ready for downstream analysis.
Optional: fill missing numeric values before the analysis runs. Leave as default to keep the method's own handling.
Back to Methods