Project A2 Hidden Tests

Question 1¶

Q1c¶

Regressive taxation overvalues inexpensive properties. In other words, these homes may be assessed at a higher price than they may actually be worth.

Question 3¶

Q3a¶

As the 2nd model has a higher complexity, we expect its training loss to be no worse than the training loss of the 1st model.

Q3b¶

The following code should return True

np.random.seed(1337)
tr_m2, te_m2 = train_val_split(full_data)
X_tr_m2, Y_tr_m2 = feature_engine_pipe(tr_m2, m2_pipelines, 'Log Sale Price')
X_te_m2, Y_te_m2 = feature_engine_pipe(te_m2, m2_pipelines, 'Log Sale Price')
# Test 1
bool(np.isclose(Y_tr_m2.sum(), 1643505.2095877344) and np.isclose(Y_te_m2.sum(), 412085.5255227775))
# Test 2
bool(np.isclose(Y_train_m1.sum(), Y_tr_m2.sum()) and np.isclose(Y_valid_m1.sum(), Y_te_m2.sum()))

Question 4¶

Q4b¶

We see that properties with a lower sale price typically have negative residuals, which means their predicted worth/sale price is overvalued.