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. You are not given a corresponding test-output.csv, instead your objective is to generate a file that predicts the outcomes for these test examples / cases and upload them here on Kaggle. Your solution will then be compared against (a subset of) the test outcomes for evaluation. You can use guess-output-common.csv and guess-output-stratified.csv to see what an example submission file should look like. These two files are baseline cases - you can even upload them to make sure your Kaggle account is working correctly!

See covtype.info 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.