Basic Usage

This page covers the basics of using the library.

Installation

Install the library with the pip package manager.

pip install sklearn-nature-inspired-algorithms

Usage

Usage is similar to scikit-learn’s GridSearchCV. The following example assumes that you are already familiar with the basics of scikit-learn and have it installed.

from sklearn_nature_inspired_algorithms.model_selection import NatureInspiredSearchCV
from sklearn.ensemble import RandomForestClassifier

param_grid = {
   'n_estimators': range(20, 100, 20),
   'max_depth': range(2, 40, 2),
   'min_samples_split': range(2, 20, 2),
   'max_features': ["auto", "sqrt", "log2"],
}

clf = RandomForestClassifier(random_state=42)

nia_search = NatureInspiredSearchCV(
   clf,
   param_grid,
   algorithm='hba', # hybrid bat algorithm
   population_size=50,
   max_n_gen=100,
   max_stagnating_gen=10,
   runs=3,
   random_state=None, # or any number for reproducible results
)

nia_search.fit(X_train, y_train)

After fitting NatureInspiredSearchCV, you can retrieve the best parameters the same way you would with GridSearchCV, using the best_params_ property. You can use them to train the final model.

new_clf = RandomForestClassifier(**nia_search.best_params_, random_state=42)

If you would like to plot the optimization process, you can use one of the built-in functions described in Visualization of Optimization Results.

Note

The fitness of each individual in the optimization process is defined by your estimator’s score method or by the scoring parameter provided to NatureInspiredSearchCV.

Examples

Jupyter notebooks with examples are available in the examples/notebooks directory.