API Reference#
Complete reference for all Gators transformers, organized by functionality.
Overview#
Gators provides 108 transformers across 11 categories, all with a consistent sklearn-compatible API.
Each transformer implements .fit() and .transform() methods and works seamlessly with Polars DataFrames.
Data Cleaning#
Quality filters, type casting, deduplication, rounding, and data quality transformations. (16 transformers)
Clippers#
Outlier detection and clipping strategies: Custom, Gaussian, IQR, MAD, and Quantile. (5 transformers)
Encoders#
Categorical encoding: Binary, CatBoost, Count, Hash, LeaveOneOut, OneHot, Ordinal, RareCategory, Target, WOE. (10 transformers)
Feature Generation#
Create numeric features: polynomial, ratios, Fourier, group aggregations, rolling windows, and custom rules. (21 transformers)
String Features#
Extract information from text: length, patterns, n-grams, TF-IDF, regex extraction, fuzzy similarity, and text statistics. (19 transformers)
DateTime Features#
Temporal feature engineering: cyclic encoding, holidays, business hours, time windows, and date differences. (8 transformers)
Imputers#
Handle missing values with mean, median, mode, constant, KNN, iterative, or group-based strategies. (6 transformers)
Discretizers#
Bin continuous variables using equal-width, quantile, geometric, k-means, tree-based, or custom strategies. (7 transformers)
Scalers#
Normalize and transform features: Standard, MinMax, Robust, BoxCox, Yeo-Johnson, Log1p, and more. (9 transformers)
Pipeline#
Chain multiple transformers together for streamlined preprocessing workflows.
Feature Selection#
Select important features using Pearson correlation, Information Value, mutual information, permutation importance, and Population Stability Index.
ONNX Export#
Export any fitted Pipeline or single transformer to a validated ONNX graph for low-latency, language-agnostic inference.
from gators.onnx_converters import pipeline_to_onnx, to_onnx_graph
# Export a complete pipeline
onnx_model = pipeline_to_onnx(fitted_pipeline)
# Export a single transformer
onnx_model = to_onnx_graph(fitted_transformer)
All transformer classes are covered by ONNX converters, with the exception of a small set of transformers whose operations have no ONNX-representable equivalent (e.g. window/partition aggregations, fuzzy string distance, and variable-length tokenization) - see each transformer’s docstring for ONNX support notes.