Deep learning models were selected for the case data model project. Each categorical input has a separate embedding layer trained from scratch. At the end of training the input layer produces Dell specific embeddings for all words. This solution relies on Long Short-Term Memory (LSTM) and skip connections.
