WGU C951 TASK 2: MACHINE
LEARNING FOR EMAIL SPAM
FILTERING
1. Which metric is most crucial to optimize if the project’s primary goal is to avoid moving
legitimate business emails to the spam folder?
A. Recall
B. Accuracy
C. Precision
D. L2 Regularization
Answer: C
Conceptual Explanation: Precision measures the ratio of true positives to the total
predicted positives. Improving precision minimizes False Positives (legitimate emails
marked as spam).
2. Why is the Multinomial Naive Bayes classifier frequently chosen for spam filtering tasks?
A. It effectively handles non-linear relationships between words.
B. It accounts for the order of words in a sentence.
C. It automatically eliminates the need for feature scaling.
,D. It is computationally efficient and performs well with high-dimensional discrete
features.
Answer: D
Conceptual Explanation: Naive Bayes is computationally fast and highly effective for text
classification where feature dimensions (vocabulary) are large but features are discrete.
3. In the context of text preprocessing, what is the main difference between stemming and
lemmatization?
A. Lemmatization considers the context and part of speech, whereas stemming follows
heuristic rules.
B. Stemming uses a dictionary to find the root, while lemmatization chops off characters.
C. Stemming is more computationally expensive than lemmatization.
D. Lemmatization always results in a shorter word than stemming.
Answer: A
Conceptual Explanation: Lemmatization uses morphological analysis to return the
dictionary base form (lemma), while stemming often just trims suffixes.
4. What problem does Laplace Smoothing solve in a Naive Bayes spam filter?
A. The curse of dimensionality.
B. Overfitting to the training set.
C. Gradient vanishing in deep networks.
, D. Zero-frequency problems for words not in the training vocabulary.
Answer: D
Conceptual Explanation: Laplace smoothing adds a small constant to the frequency
counts to ensure that the probability of a word is never zero, preventing the entire product
of probabilities from zeroing out.
5. Which feature engineering technique represents a document by the frequency of words
while penalizing words that appear across many documents?
A. TF-IDF
B. Word2Vec
C. One-Hot Encoding
D. Principal Component Analysis
Answer: A
Conceptual Explanation: TF-IDF (Term Frequency-Inverse Document Frequency)
highlights words that are unique to specific documents while downweighting common
words like ‘the’ or ‘and’.
6. What is a major limitation of the Bag-of-Words (BoW) model in sensitive spam detection?
A. It requires a GPU to compute vector distances.
B. It ignores the semantic meaning and the order of words.
C. It cannot handle numerical data.
LEARNING FOR EMAIL SPAM
FILTERING
1. Which metric is most crucial to optimize if the project’s primary goal is to avoid moving
legitimate business emails to the spam folder?
A. Recall
B. Accuracy
C. Precision
D. L2 Regularization
Answer: C
Conceptual Explanation: Precision measures the ratio of true positives to the total
predicted positives. Improving precision minimizes False Positives (legitimate emails
marked as spam).
2. Why is the Multinomial Naive Bayes classifier frequently chosen for spam filtering tasks?
A. It effectively handles non-linear relationships between words.
B. It accounts for the order of words in a sentence.
C. It automatically eliminates the need for feature scaling.
,D. It is computationally efficient and performs well with high-dimensional discrete
features.
Answer: D
Conceptual Explanation: Naive Bayes is computationally fast and highly effective for text
classification where feature dimensions (vocabulary) are large but features are discrete.
3. In the context of text preprocessing, what is the main difference between stemming and
lemmatization?
A. Lemmatization considers the context and part of speech, whereas stemming follows
heuristic rules.
B. Stemming uses a dictionary to find the root, while lemmatization chops off characters.
C. Stemming is more computationally expensive than lemmatization.
D. Lemmatization always results in a shorter word than stemming.
Answer: A
Conceptual Explanation: Lemmatization uses morphological analysis to return the
dictionary base form (lemma), while stemming often just trims suffixes.
4. What problem does Laplace Smoothing solve in a Naive Bayes spam filter?
A. The curse of dimensionality.
B. Overfitting to the training set.
C. Gradient vanishing in deep networks.
, D. Zero-frequency problems for words not in the training vocabulary.
Answer: D
Conceptual Explanation: Laplace smoothing adds a small constant to the frequency
counts to ensure that the probability of a word is never zero, preventing the entire product
of probabilities from zeroing out.
5. Which feature engineering technique represents a document by the frequency of words
while penalizing words that appear across many documents?
A. TF-IDF
B. Word2Vec
C. One-Hot Encoding
D. Principal Component Analysis
Answer: A
Conceptual Explanation: TF-IDF (Term Frequency-Inverse Document Frequency)
highlights words that are unique to specific documents while downweighting common
words like ‘the’ or ‘and’.
6. What is a major limitation of the Bag-of-Words (BoW) model in sensitive spam detection?
A. It requires a GPU to compute vector distances.
B. It ignores the semantic meaning and the order of words.
C. It cannot handle numerical data.
