train-features.csv contains the features of the examples / cases, separated by commas, that you will use to train your classifiers. You must associate the features found in each row / line with the output found in the corresponding row / line in train-output.csv to create a full case.

test-features.csv contains the features of the examples / cases that you will need to predict using your trained model.

See train.names for header descriptions of the train and test features. This will be helpful in understanding how you may need to normalize, bin, or otherwise update your data to better suit your classifiers.

You can use test-submission.demo to see what an example submission file should look like. This file is a baseline case - you can even upload it to make sure your Kaggle account is working correctly!