Preprocessing Help

Impute missing values, normalise features and handle outliers. Produces a derived dataset for downstream use.

Requires: One dataset with numeric columns.

Form fields accepted by this method's POST request

analysis_type=preprocessing, dataset_id, min_col_fraction, missing_strategy, normalization, outlier_handling, outlier_method, target_col

No dataset yet? Upload a file.
If set, the target is preserved in the derived dataset but excluded from feature transforms.
Parameters & Options
For IQR/Z-Score: a row is only flagged when it is outlying in at least this fraction of numeric columns (default 10%). Prevents mass row deletion on wide datasets.
Back to Methods