Katkat
Back

Build Naive Bayes spam classifier with F1

DisneyMachine Learning Engineer · Technical Screen
MediumMachine Learning EngineerTechnical ScreenMachine Learning

Last updated

This question evaluates understanding of probabilistic text classification with Naive Bayes, text preprocessing and feature extraction, and use of the F1 score for performance measurement in a binary spam detection task.

You are given a text classification dataset for spam detection (binary labels: spam vs not_spam) in a Jupyter notebook environment.

Task

  1. Preprocess the text (basic cleaning/tokenization is sufficient).
  2. Convert text to features suitable for Naive Bayes (e.g., bag-of-words or TF-IDF).
  3. Train a Naive Bayes classifier.
  4. Evaluate the model using F1 score (clearly state whether it is the F1 for the positive class or a specific averaging scheme).
  5. Run the trained model on a few test examples and show predicted labels (and optionally probabilities).

Constraints / Notes

  • The dataset may be class-imbalanced.
  • You should avoid data leakage (fit text vectorizer only on training data).
  • You may choose reasonable train/validation splitting if only one labeled set is provided.