CODE & RESULTS

Logistic Regression achieved a slightly higher accuracy (51.11%) compared to Multinomial Naive Bayes (49.94%).

Logistic Regression correctly predicted more samples from Class 1 (high severity) but misclassified many Class 0 samples, indicating bias toward predicting high severity.

Multinomial Naive Bayes produced nearly equal predictions for both classes, reflecting balanced but weaker discriminatory power.

Both models show significant overlap in misclassifications, suggesting difficulty in distinguishing between mild and severe anxiety from the features used.

Overall, Logistic Regression was marginally better at classifying samples, though both models struggled due to subtle or nonlinear patterns in the data.

Logistic Regression outperformed Multinomial Naive Bayes slightly, achieving an accuracy of around 51% versus 49.9% for NB.

Both models hover around 50% accuracy, indicating that neither model is confidently separating the classes performance is close to random guessing.

Logistic Regression may better capture linear relationships between the features and target variable, giving it a slight edge.

Multinomial Naive Bayes, which relies on feature independence and discrete count-based assumptions, likely underperforms due to the nature of the dataset (continuous and possibly correlated features).

This visualization highlights the need for either feature engineering, better preprocessing, or trying more flexible models (like decision trees or ensemble methods) for improved performance.