A simple, fast classification model that combines the features with learned weights and outputs a probability.

ExampleOn our split it reaches about 80% test accuracy, against 73.5% for the baseline.