WGU C951 TASK 1: ADVANCED
MACHINE LEARNING FOR EMAIL
SPAM FILTERING
1. When preparing the dataset for a spam filter, which technique is most effective for
addressing the extreme class imbalance often found in real-world email traffic?
A. Applying Synthetic Minority Over-sampling Technique (SMOTE) on the feature space
B. Simple random oversampling of the minority class
C. Increasing the learning rate of the optimization algorithm
D. Using a standard 50/50 split for training and testing
Answer: A
Conceptual Explanation: SMOTE generates synthetic examples rather than just
duplicating records, which helps the model generalize better for the minority (spam) class
in an imbalanced dataset.
2. In the context of Naive Bayes for spam detection, what is the primary purpose of applying
Laplace smoothing?
A. To account for words in the test set that never appeared in the training set
B. To reduce the computational complexity of the probability calculations
,C. To normalize the length of different email messages
D. To handle the high correlation between different features
Answer: A
Conceptual Explanation: Laplace smoothing (or add-one smoothing) prevents the
posterior probability from becoming zero when a word in the test set was not present in
the training data.
3. Which evaluation metric is most critical for an email service provider to minimize if they
want to avoid losing legitimate business communications?
A. False Negative Rate
B. Recall
C. Accuracy
D. False Positive Rate
Answer: D
Conceptual Explanation: A False Positive occurs when a legitimate email (ham) is
incorrectly classified as spam. Minimizing this rate is vital for user trust.
4. When utilizing TF-IDF for feature extraction, what does the ‘IDF’ component specifically
help to penalize?
A. Words that appear frequently in a single document but rarely across the corpus
B. Misspelled words commonly found in spam subject lines
, C. Short emails that do not contain enough semantic information
D. Words that appear frequently across all documents in the corpus
Answer: D
Conceptual Explanation: Inverse Document Frequency (IDF) reduces the weight of terms
that occur very frequently across the entire corpus, such as ‘the’ or ‘and’, which are less
discriminative.
5. Why might a Support Vector Machine (SVM) with a Radial Basis Function (RBF) kernel
outperform a Linear SVM in spam detection?
A. The decision boundary between spam and ham is often non-linear
B. The RBF kernel requires less memory for large datasets
C. Linear SVMs cannot handle sparse matrices generated by Bag-of-Words
D. RBF kernels are inherently faster to train on text data
Answer: A
Conceptual Explanation: Spam features often have complex, non-linear interactions that
an RBF kernel can capture by mapping features into a higher-dimensional space.
6. In a spam filtering project proposal, which concept describes the risk of the model
performance degrading over time as spammers change their tactics?
A. Overfitting
B. Data Leakage
MACHINE LEARNING FOR EMAIL
SPAM FILTERING
1. When preparing the dataset for a spam filter, which technique is most effective for
addressing the extreme class imbalance often found in real-world email traffic?
A. Applying Synthetic Minority Over-sampling Technique (SMOTE) on the feature space
B. Simple random oversampling of the minority class
C. Increasing the learning rate of the optimization algorithm
D. Using a standard 50/50 split for training and testing
Answer: A
Conceptual Explanation: SMOTE generates synthetic examples rather than just
duplicating records, which helps the model generalize better for the minority (spam) class
in an imbalanced dataset.
2. In the context of Naive Bayes for spam detection, what is the primary purpose of applying
Laplace smoothing?
A. To account for words in the test set that never appeared in the training set
B. To reduce the computational complexity of the probability calculations
,C. To normalize the length of different email messages
D. To handle the high correlation between different features
Answer: A
Conceptual Explanation: Laplace smoothing (or add-one smoothing) prevents the
posterior probability from becoming zero when a word in the test set was not present in
the training data.
3. Which evaluation metric is most critical for an email service provider to minimize if they
want to avoid losing legitimate business communications?
A. False Negative Rate
B. Recall
C. Accuracy
D. False Positive Rate
Answer: D
Conceptual Explanation: A False Positive occurs when a legitimate email (ham) is
incorrectly classified as spam. Minimizing this rate is vital for user trust.
4. When utilizing TF-IDF for feature extraction, what does the ‘IDF’ component specifically
help to penalize?
A. Words that appear frequently in a single document but rarely across the corpus
B. Misspelled words commonly found in spam subject lines
, C. Short emails that do not contain enough semantic information
D. Words that appear frequently across all documents in the corpus
Answer: D
Conceptual Explanation: Inverse Document Frequency (IDF) reduces the weight of terms
that occur very frequently across the entire corpus, such as ‘the’ or ‘and’, which are less
discriminative.
5. Why might a Support Vector Machine (SVM) with a Radial Basis Function (RBF) kernel
outperform a Linear SVM in spam detection?
A. The decision boundary between spam and ham is often non-linear
B. The RBF kernel requires less memory for large datasets
C. Linear SVMs cannot handle sparse matrices generated by Bag-of-Words
D. RBF kernels are inherently faster to train on text data
Answer: A
Conceptual Explanation: Spam features often have complex, non-linear interactions that
an RBF kernel can capture by mapping features into a higher-dimensional space.
6. In a spam filtering project proposal, which concept describes the risk of the model
performance degrading over time as spammers change their tactics?
A. Overfitting
B. Data Leakage
