Stacking Regressor
import pandas as pd
import numpy as np
import seaborn as sns
sns.get_dataset_names()
mpg = sns.load_dataset("mpg")
mpg
mpg = mpg.drop('name', axis=1)
mpg = pd.get_dummies(mpg)
mpg.head(10)
SEE ALL NULL VLAUES
pd.DataFrame(mpg.isnull().sum().sort_values(ascending=False))
mpg['horsepower'].fillna(mpg['horsepower'].mean(), inplace=True)
pd.DataFrame(mpg.isnull().sum().sort_values(ascending=False))
X = mpg.drop('mpg', axis=1)
y = mpg['mpg']
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=19)
from sklearn.linear_model import LinearRegression
lr = LinearRegression()
lr.fit(X_train, y_train)
y_pred = lr.predict(X_test)
from sklearn.metrics import mean_absolute_error, mean_squared_error, r2_score
mean_squared_error(y_test, y_pred)
from sklearn.ensemble import RandomForestRegressor
rfr = RandomForestRegressor(random_state=13)
rfr.fit(X_train, y_train)
y_pred = rfr.predict(X_test)
mean_squared_error(y_test, y_pred)
from sklearn.linear_model import Ridge
ridge = Ridge()
param_grid = {
'alpha': [0.05, 0.1, 0.3, 1, 3, 5, 10, 15, 30, 50, 75],
}
from sklearn.model_selection import GridSearchCV
ridge_cv = GridSearchCV(estimator=ridge, param_grid=param_grid, cv=5, scoring='neg_mean_squared_error', n_jobs=-1)
ridge_cv.fit(X_train, y_train)
ridge_cv.best_estimator_
mean_squared_error(y_test, y_pred)
from sklearn.ensemble import GradientBoostingRegressor
gbr = GradientBoostingRegressor()
gbr.fit(X_train, y_train)
y_pred = gbr.predict(X_test)
mean_squared_error(y_test, y_pred)
from sklearn.ensemble import StackingRegressor
estimators = [
('lr', lr),
('ridge', ridge_cv.best_estimator_),
('gbr', gbr),
]
sr = StackingRegressor(
estimators=estimators,
final_estimator=rfr
)
sr.fit(X_train, y_train)
y_pred = sr.predict(X_test)
mean_squared_error(y_test, y_pred)
voting classifier
from sklearn.ensemble import VotingRegressor
vc = VotingRegressor([('lr', lr),
('rfr', rfr),
], weights=[1,2])
vc.fit(X_train, y_train)
y_pred = vc.predict(X_test)
mean_squared_error(y_test, y_pred)
sr2 = StackingRegressor(
estimators=estimators,
final_estimator=vc
)
sr2.fit(X_train, y_train)
y_pred = sr2.predict(X_test)
mean_squared_error(y_test, y_pred)
hyperparamater tuning
from sklearn.svm import SVR
base_regressors = [
('ridge', ridge_cv.best_estimator_),
('gbr', gbr),
('svr', SVR(C=1.0, kernel='linear')),
('random_forest', RandomForestRegressor()),
]
stacking_regressor = StackingRegressor(
estimators=base_regressors,
final_estimator=Ridge(alpha=1.0)
)
param_grid = {
'random_forest__n_estimators': [50, 100, 250],
'svr__C': [0.1, 1.0, 10.0],
'final_estimator__alpha': [0.1, 1.0, 10.0],
}
from sklearn.model_selection import RandomizedSearchCV
random_search = RandomizedSearchCV(stacking_regressor, param_grid, n_iter=10, cv=3, scoring='neg_mean_squared_error', n_jobs=-1)
random_search.fit(X_train, y_train)
y_pred = random_search.predict(X_test)
mean_squared_error(y_test, y_pred)
random_search.best_params_
