Study shows 96.5% accurate spam filter missed 58 spam messages

A developer trained a logistic regression model on a dataset of 5,572 emails to create a spam filter. The model achieved 96.5% test accuracy, improving significantly over a baseline that always predicted 'ham'. However, its confusion matrix revealed it correctly identified only 174 out of 232 spam messages, letting 58 through. The developer noted the model prioritizes precision over recall for spam, making it safer for avoiding false positives. They also corrected their project documentation after discovering a discrepancy between their written description and the actual code.