Pipelines, Saving and Loading models

Pipelines

from sklearn.pipeline import Pipeline
operations = [('scaler',scaler),('knn',knn)]
pipe = Pipeline(operations)
k_values=list(range(1,20))
param_grid = {'knn__n_neighbors':k_values}

from sklearn.model_selection import GridSearchCV
full_cv_classifier = GridSearchCV(pipe,param_grid,cv=5,scoring='accuracy')
full_cv_classifier.fit(scaled_X_train,y_train)

Saving and loading models

Saving:

from joblib import dump
dump(final_model,'trained_model.joblib')
dump(list(X.columns),'colnames.pkl')

Loading:

from joblib import load
loaded_model = load('trained_model.joblib')
new_columns = joblib.load('colnames.pkl')

• sometimes it is useful to save the column names for the dataframe. This is not strictly needed though.