Metrics - prediction task is binary classification task and the authors use AUC (area under curve) to measure the performance. Approaches - their created model is compared to previous models, namely, DKT (deep knowledge tracing), DKT+, DKVMN(dynamic key value memory network). Model training and parameter selection the dataset was divided into train/test sets in the following way: 80% goes to train and 20% to test. Different size of hidden state have been tried: {50, 100, 150, 200}. The model was implemented in tensorflow and adam optimizer with 0.001 was utilized.