additive-sparse-boost-regression
A Python Package for a Sparse Additive Boosting Regressor
https://github.com/thesis-jdgs/additive-sparse-boost-regression
Science Score: 44.0%
This score indicates how likely this project is to be science-related based on various indicators:
-
✓CITATION.cff file
Found CITATION.cff file -
✓codemeta.json file
Found codemeta.json file -
✓.zenodo.json file
Found .zenodo.json file -
○DOI references
-
○Academic publication links
-
○Academic email domains
-
○Institutional organization owner
-
○JOSS paper metadata
-
○Scientific vocabulary similarity
Low similarity (9.8%) to scientific vocabulary
Keywords
Repository
A Python Package for a Sparse Additive Boosting Regressor
Basic Info
Statistics
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
- Releases: 0
Topics
Metadata Files
README.md
AdditiveSparseBoostRegressor
This repository holds the implementation for an additive regressor, i.e.:
math
y\approx f(x) = \beta + \sum_{i=1}^M f_i(x_i)
Where $f_i$ functions with zero-mean and $\beta$ is the intercept, both parameters are estimated by minimizing the squared error loss function.
The functions $f_i$ are either a piecewise constant function (trained as ensemble of decision trees), or the trivial zero-function, that is to say, some features are ignored, making the regressor sparse.
This is achieved by combining the boosting algorithm with a modification of mRMR (minimum Redundancy Maximum Relevance) feature selection at each boosting iteration.
Installation
The package can be installed with pip:
bash
pip install git+https://github.com/thesis-jdgs/additive-sparse-boost-regression.git
Note: The package is not yet available on PyPI.
Usage
The regressor is implemented in asboostreg.py,
and implements the familiar fit and predict methods from scikit-learn.:
```python from asboostreg import SparseAdditiveBoostingRegressor from sklearn.datasets import load_boston
X, y = loadboston(returnXy=True) sparsereg = SparseAdditiveBoostingRegressor( learningrate=0.01, nestimators=10000, l2regularization=2.0, maxdepth=6, rowsubsample=0.632, randomstate=0, niternochange=30, ) sparsereg.fit(X, y) ypred = sparsereg.predict(X) ```
To inspect the general characteristics of the model,
the plot_model_information method creates a plotly figure:
python
sparsereg.plot_model_information()
Which creates a figure of the iteration history and model complexity for each feature,
like the following:
To inspect the predictions for a dataset X, you can use explain(X):
python
sparsereg.explain(X)
Which creates a figure of the mean importances of each feature,
and a plot of the 1D regressor for each selected feature:
We can also decompose the predictions into the additive components,
with the contribution_frame method:
python
sparsereg.contribution_frame(X)
Which returns a pandas DataFrame with the additive components for each feature.
Owner
- Name: thesis-jdgs
- Login: thesis-jdgs
- Kind: organization
- Repositories: 1
- Profile: https://github.com/thesis-jdgs
Citation (CITATION.cff)
cff-version: 1.2.0 message: "If you use this software, please cite it as below." authors: - family-names: "Godoy Sánchez" given-names: "Johnny Godoy" orcid: "https://orcid.org/0009-0000-1202-1106" title: "additive-sparse-boost-regression" version: 1.0.0 date-released: 2023-05-21 url: "https://github.com/thesis-jdgs/additive-sparse-boost-regression"
GitHub Events
Total
Last Year
Dependencies
- attrs *
- category_encoders *
- numpy *
- pandas *
- plotly *
- scikit-image *
- scikit-learn *
- interpret *
- jupyter *
- xgboost *
- line.strip *