Machine Learning — Forecasting Using Regression Models (part3)

Model Engineering — select and evaluate linear / non-linear models

from unsplash

From previous post, we can know that:

• Knowing that this dataset is non-linear, using linear regression will not return meaningful results. However, it is always good to have a baseline. Below linear regressions are tryout to give a baseline.
• Use non-linear regression algorithm MLP Regressor and Random Forest Regressor.
• Use GridSearchCV to split the dataset to 10 folds.

— — — — — —USE OF LINEAR MODELS— — — — — —

Use of Linear Regression, SGDRegressor and Support Vector Regression:

from sklearn.linear_model import LinearRegression, SGDRegressor
def fit_model_cross_validate(algorithm, parameters, X_train, y_train, X_test, y_test):
 gs = GridSearchCV(algorithm, parameters, cv=10, verbose=10)
 gs.fit(X_train, y_train)
 print(‘Best parameters:’, gs.best_params_)
 print(‘Score on test set:’, gs.score(X_test, y_test))
 # return both the model and the predictions
 return gs, gs.predict(X_test)
def evaluate_model(X_test, y_test):
 print(‘MSE’, mean_squared_error(X_test, y_test))
 
def check_goodness_of_fit(model, X_train, y_train, X_test, y_test)…