Task 4 · Natural language processing
Email Spam Detection
Cleans each message, turns it into TF-IDF features and classifies it with Multinomial Naive Bayes, trained on 5,171 SMS messages. Recall matters here: every missed spam lands in someone's inbox.
98.8%accuracy
96.1%spam precision
94.7%spam recall
0.954spam F1
Try it: is this message spam?
Examples:
Verdict
–
Words driving the decision
The browser version uses the notebook's pipeline without word stemming so it can run in JavaScript; it scores 99.1% accuracy on the same test set.
