Basic Usage¶
This page covers the basics of using the library.
Installation¶
Install the library with pip package manager.
pip install sklearn-nature-inspired-algorithms
Usage¶
The usage is similar to using sklearn’s GridSearchCV. In the following example, we assume that you are already familiar with the basics of the scikit-learn library. Also you need to 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 if you want same results on each run
)
nia_search.fit(X_train, y_train)
After you have fit the NatureInspiredSearchCV, you can retrieve the best parameters in the same way as with GridSearchCV using the best_params_ property. You can use them to train the final model.
If you would like to plot the optimization process you can use one of the built-in functions described in Visualisation of Optimization Results.
Note
The fitness of each invididual in the optimization process is defined by the score method of your estimator or the scoring parameter provided to the NatureInspiredSearchCV.