Question 1

You are building a predictive model using a dataset that contains numerous variables. Upon inspection, you notice a significant number of them are highly correlated. Which of the following actions is the MOST appropriate to address potential issues with irrelevant or redundant variables before proceeding with model building?

Options :

A : Remove all variables with a correlation coefficient above 0.8 to eliminate redundancy.

B : Use a dimensionality reduction technique like Principal Component Analysis (PCA).

C : Scale all variables to have zero mean and unit variance without considering their correlations.

D : Increase the model complexity to account for the additional information in the correlated variables.

Answer: B

Question 2

You are building an artificial neural network model to predict the probability that a customer will respond to a marketing campaign. The target variable is binary, indicating a '1' for a response and '0' for no response. Which combination of activation function and error function would be most appropriate for the output layer of this neural network?

Options :

A :

Activation function: Hyperbolic tangent (tanh), Error function: Cross-entropy

B :

Activation function: Softmax, Error function: Mean squared error (MSE)

C :

Activation function: Sigmoid, Error function: Cross-entropy

D :

Activation function: ReLU (Rectified Linear Unit), Error function: Mean absolute error (MAE)

Answer: C

Question 3

When preparing data for a predictive modeling project, a data scientist notices that the categorical variable 'payment_type' with four categories ('credit card', 'debit card', 'paypal', 'other') exhibits a high degree of variability in the outcome variable (purchase amount). To improve the model's predictive accuracy, what strategy can the data scientist use to handle the 'payment_type' variable?

Options :

A :

Group all categories into a single 'payment_type' indicator variable.

B :

Perform one-hot encoding on 'payment_type' to create a binary indicator for each category.

C :

Collapse 'paypal' and 'other' into a single category if they have similar effects on the purchase amount.

D :

Treat 'payment_type' as a continuous variable to simplify the model.

Answer: B

Question 4

You have built a logistic regression model using PROC LOGISTIC to predict the probability of customer churn based on several predictors. You want to evaluate the performance of the model by analyzing the Receiver Operating Characteristic (ROC) curve. To score a validation dataset and produce the ROC curve, which statement correctly implements the OUTROC option in the SCORE statement?

Options :

A :

SCORE DATA=validation OUT=scored_results OUTROC=roc_curve;

B :

PROC LOGISTIC DATA=training; SCORE OUTROC=roc_curve; RUN;

C :

PROC LOGISTIC DATA=training; SCORE DATA=validation OUTROC=roc_curve; RUN;

D :

SCORE DATA=validation OUT=scored_results; MODEL OUTROC=roc_curve;

Answer: C

Question 5

As a data scientist, you have built three predictive models to forecast the risk of a rare event occurring within a patient group. To evaluate these models, you have computed several fit statistics. Consider the following statistics for the models: Model A: - BIC: 182 - AIC: 175 - KS: 0.65 - Brier Score: 0.12 Model B: - BIC: 185 - AIC: 178 - KS: 0.60 Brier Score: 0.11 Model C: - BIC: 180 - AIC: 182 - KS: 0.80 - Brier Score: 0.09 Assuming that the most important criteria for model selection is the prediction accuracy of the rare event and considering the disease is highly imbalanced, which model should you recommend?

Options :

A :

Model A

B :

Model B

C :

Model C

D :

More information is needed regarding the internal structure of the models and the distribution of the data.

Answer: C

Confidently Practice Online with Free A00-225 Exam Cram

Buying Options: