There are specifically 3 files with data columns linked by unique keys. The keys are spread across training and test set. The implementation uses several parametric as well as non-parametric binary classification algorithms in addition to using a neural network with dynamic(based on number of input variables) hidden layer inputs