This document compares several machine learning algorithms for a binary classification problem using a census dataset:
1. It builds logistic regression, decision tree, random forest, and boosted tree models on a 80% training set and evaluates their performance on a 10% test set.
2. Tuning is performed on decision tree and random forest models which improves their AUC.
3. The best performing models are boosted trees with an AUC of 0.922 and logistic regression with an AUC of 0.91, as evaluated on the held-out test set.