Modul 3 Machine Learning

The function above will calculate an evaluation metric, which is

• R-squared

• Root mean squared error

• Mean squared error

• Mean absolute error

• Mean absolute percentage error

Given the linear regression model equation that used to predict used car prices (in Rupiah).

The correct conclusion based on the equation above is

• The further the mileage of a car is, the more expensive the used selling price will be

• Passenger capacity affects the selling price of a used car

• A used car produced in 2000 with a mileage of 150,000 kilometers is predicted to have a price of Rp. 255,300,000

• Cars produced in 2008 with a range of 100,000 kilometers will be predicted to be more expensive than cars produced in 2012 with a range of 150,000 kilometers

• The model above cannot be used to predict a used car with the odometer (mileage indicator) in unit of miles

Which is the correct statement about tuning the k parameter in K-Nearest Neighbors for the classification case?

• changing the value of k will not affect the number of iterations performed by the algorithm

• large k values always make the model overfitting

• if the value of k is equal to the number of data points in the training set, all data will be classified into the same class, which is the majority class

• increasing the value of k tends to make the algorithm work faster

• the best k value is the k value that is able to produce the smallest RMSE

The following evaluation metrics are commonly used for regression cases. Which metric is NOT for the regression problem?

• MAE

• RMSE

• F1 Score

• R2 Score

• All of these metrics are for regression

Look at the following table of evaluation scores from several candidates of models.

The correct statement about the table above is ….

• Model No.2 is overfitting

• The performance of Model No.2 will definitely improve by reducing the number of features

• Model No.3 is more likely to be selected as the best model candidate than Model No.1 or Model No.2

• The number of features does not have any impact on the F1 score

• The performance of the Model No.1 will definitely improve by increasing the number of features

The role of alpha value in lasso regression is ….

Choose the correct answer.

• To modify the power order of the independent variable

• To determine the number of independent variables given a penalty

• To determine the degree of penalty assigned to the independent variable

• To be the multiplier coefficient for each independent variable

• None of the other options is correct

Among the several variable examples below, one-hot encoding should be avoided to be applied to

• Driving License Card Number

• Education level

• Marital status

• Type of work

Look at the following code.

If X_train[ctg_cols] has no missing values, X_train_ctg_encoded will has …. columns

• 6

• 12

• 4

• 2

• 5

Among the options below, which is the weakness of the model-based feature selection?

• It can not capture the interaction between independent variables

• It takes time to train a model first

• Importance of a feature does not depend on other features

• It can’t be used for categorical features

• None of the other options are correct

If "YES" is assumed to be a positive class, then what is the accuracy of the confusion matrix above?

• 0.975

• 0.889

• 0.845

• 0.6

• 0.333