BaseRush
baserush
Stable base modeling made quick and easy!
baserush is an easy-to-use regression pipeline for preprocessing, optimizing, and summarizing machine learning models within the scikit-learn framework. This package is ideal for efficiently building and comparing stable models from different model types.
Supported Model Types
Linear Models
- LinearRegression
- Lasso
- Ridge
- SGDRegressor
Neighbors Models
- KNeighborsRegressor
- RadiusNeighborsRegressor
CaRT Models
- DecisionTreeRegressor
- ExtraTreeRegressor
Ensemble Models
- RandomForestRegressor
- GradientBoostingRegressor
- ExtraTreesRegressor
- RandomTreesEmbedding
Package Modules
preprocess: missing values, skewness, standardization, and categorical transformationsoptimize: automatic feature selection; hyperparameter analysissummary: training and validation R-Squared, stability tools; model-specific outputs
preprocessing Features
simputermakes it simple to flag and impute missing values.- Quickly alleviate skewness with
transtorm. - Use
simple_scalerto seamlessly standardize features. - Efficiently prepare categorical data for modeling with
catcoder.
optimize-ation Features
- Base modeling made easy with
quick_lm(with automated feature selection)quick_tree, (includes very fast automated hyperparameter tuning)quick_neighbors, (automatically tunes n neighbors)
- Use
tuning_resultsto analyze the top n-models after hyperparameter tuning with GridSeachCV | RandomizedSearchCV.
summary Features
lr_summary, tree_summary, and knn_summary
- Automatically instantiate customizable training and validation sets.
- Generate a dataset of model summaries for easy comparison, including:
- Model Name
- Model Class
- Model Type
- R-Squared (Training Set)
- R-Squared (Validation Set)
- Train-Test Gap
- Model-Specific:
- Model Coefficients
- Feature Importance
- Hyperparameter Values
Installation
Install using pip:
pip install baserush
Example Usage
print("Examples coming soon.")
License
MIT License. See LICENSE for details.