Two-Layer Additive Risk Model - A fully transparent global prediction model for the FICO dataset
For a new individual,
a personalized local model is generated, which is simpler than the global model but is still consistent with it;
nearest neighbors are generated with corresponding learned distance metric.