Naive Bayes Q&A
20 Core Questions
Interview Prep
Naive Bayes: Interview Q&A
Short questions and answers on Naive Bayes: Bayes theorem, independence assumption and popular variants for text and numeric data.
Bayes
Independence
Text
Gaussian
1
What is Naive Bayes in simple terms?
âš¡ Beginner
Answer: Naive Bayes is a probabilistic classifier that applies Bayes’ theorem assuming features are conditionally independent given the class.
2
Why is it called “naive�
âš¡ Beginner
Answer: Because it makes the strong and usually unrealistic assumption that all features are independent given the class.
3
State Bayes’ theorem briefly.
âš¡ Beginner
Answer: Bayes’ theorem says \(P(A|B) = \frac{P(B|A)P(A)}{P(B)}\), relating posterior, likelihood, prior and evidence.
4
What are some common variants of Naive Bayes?
📊 Intermediate
Answer: Popular variants include Gaussian, Multinomial and Bernoulli Naive Bayes.
5
Which Naive Bayes variant is commonly used for text classification?
âš¡ Beginner
Answer: Multinomial Naive Bayes is widely used for bag‑of‑words and TF‑IDF text features.
6
When do you use Gaussian Naive Bayes?
📊 Intermediate
Answer: Gaussian NB is used for continuous numeric features that are roughly normally distributed within each class.
7
Why is Naive Bayes fast to train?
âš¡ Beginner
Answer: Training mainly involves counting feature occurrences per class and computing simple statistics, with no iterative optimization.
8
What is Laplace (add-one) smoothing and why is it used?
📊 Intermediate
Answer: Laplace smoothing adds a small constant (e.g., 1) to counts to avoid zero probabilities for unseen features in a class.
9
How does Naive Bayes make a classification decision?
âš¡ Beginner
Answer: It computes the posterior probability for each class given the features and picks the class with the highest posterior.
10
What are the main strengths of Naive Bayes?
âš¡ Beginner
Answer: It is simple, fast, robust to irrelevant features and works surprisingly well for high-dimensional text tasks.
11
What are some weaknesses of Naive Bayes?
📊 Intermediate
Answer: It can perform poorly when feature independence is badly violated or when interactions between features are crucial.
12
Are Naive Bayes probability estimates well calibrated?
🔥 Advanced
Answer: They are often poorly calibrated and over‑confident, even when classification accuracy is good.
13
Why can Naive Bayes still work well even if the independence assumption is false?
🔥 Advanced
Answer: Because often the classification decision only needs a rough ranking of posteriors, and Naive Bayes can get that right despite mis-specified probabilities.
14
How does class prior probability affect predictions?
📊 Intermediate
Answer: The prior P(class) biases predictions toward more probable classes, especially when evidence from features is weak.
15
How do you handle continuous features in Multinomial Naive Bayes?
🔥 Advanced
Answer: They’re typically discretized or transformed into counts (e.g., binning) before applying the multinomial model.
16
When is Naive Bayes a good baseline model?
âš¡ Beginner
Answer: It’s a strong baseline for text classification, spam filtering and document tagging, where features are word counts.
17
Does Naive Bayes work better with many or few features?
📊 Intermediate
Answer: It can handle very many features well, as long as they add independent evidence and counts are reliable.
18
How do you evaluate a Naive Bayes classifier?
âš¡ Beginner
Answer: Using standard classification metrics: accuracy, precision, recall, F1, ROC‑AUC, plus confusion matrices.
19
Give a famous real-world use case of Naive Bayes.
âš¡ Beginner
Answer: Email spam filtering is a classic application where Naive Bayes was very successful.
20
What is the key message to remember about Naive Bayes?
âš¡ Beginner
Answer: Naive Bayes is simple, fast and often effective; understand its independence assumption, smoothing and when its rough probability estimates are good enough.
Quick Recap: Naive Bayes
If you can explain Bayes’ rule, the independence assumption and why it still works for text, you’ll be ready for most Naive Bayes interview questions.