def train_and_evaluate(model, parameters):
train_n = train_x.shape[0]
train_p = train_x.shape[1]
test_n = test_x.shape[0]
test_p = test_x.shape[1]
grid_model = GridSearchCV(estimator=model(), param_grid=parameters, cv=5, n_jobs=3, verbose=1)
grid_model.fit(train_x, train_y)
print(" Results from Grid Search " )
print("\n The best estimator across ALL searched params:\n", grid_model.best_estimator_)
print("\n The best score across ALL searched params:\n", grid_model.best_score_)
print("\n The best parameters across ALL searched params:\n", grid_model.best_params_)
print()
pred_y = grid_model.predict(test_x)
print(f"RMSE: {mean_squared_error(pred_y * test_df['floor_area_sqm'], test_df['resale_price'], squared=False)}")
print(f"MSE: {mean_squared_error(pred_y * test_df['floor_area_sqm'], test_df['resale_price'], squared=True)}")
print(f"MAPE: {mean_absolute_percentage_error(pred_y * test_df['floor_area_sqm'], test_df['resale_price'])}" )
print(f"AdjR2: {1 - (1 - r2_score(test_df['resale_price'], pred_y * test_df['floor_area_sqm'])) * (test_n - 1) / (test_n - test_p - 1)}")
print()
plt.scatter(pred_y * test_df['floor_area_sqm'], test_df['resale_price'] - pred_y * test_df['floor_area_sqm'])
plt.xlabel("Predicted Price")
plt.ylabel("Residual")
plt.title("Test Residual Plot")
plt.show()
train_score_1, test_score_1 = grid_model.score(train_x, train_y), grid_model.score(test_x, test_y)
grid_model_2 = GridSearchCV(estimator=model(), param_grid=parameters, cv=5, n_jobs=3, verbose=1)
grid_model_2.fit(train_price_x, train_price_y)
print(" Results from Grid Search " )
print("\n The best estimator across ALL searched params:\n", grid_model_2.best_estimator_)
print("\n The best score across ALL searched params:\n", grid_model_2.best_score_)
print("\n The best parameters across ALL searched params:\n", grid_model_2.best_params_)
print()
pred_price_y = grid_model_2.predict(test_price_x)
print(f"RMSE: {mean_squared_error(pred_price_y, test_price_y, squared=False)}")
print(f"MSE: {mean_squared_error(pred_price_y, test_price_y, squared=True)}")
print(f"MAPE: {mean_absolute_percentage_error(pred_price_y, test_price_y)}" )
print(f"AdjR2: {1 - (1 - r2_score(test_price_y, pred_price_y)) * (test_n - 1) / (test_n - test_p - 1)}")
print()
plt.scatter(pred_price_y, test_price_y - pred_price_y)
plt.xlabel("Predicted Price")
plt.ylabel("Residual")
plt.title("Test Residual Plot")
plt.show()
train_score_2, test_score_2 = grid_model_2.score(train_price_x, train_price_y), grid_model_2.score(test_price_x, test_price_y)
print('TRAIN')
print('Using price_per_sqm:', train_score_1)
print('Using resale_price:', train_score_2)
print()
print('TEST')
print('Using price_per_sqm:', test_score_1)
print('Using resale_price:', test_score_2)
print()
return grid_model, grid_model_2
