Fig. 3

Full memory time window. A dynamical modeling framework for credit risk assessment was recently proposed by Maria Rocha Sousa et al. (Sousa & Gama, 2016) that extends the prevailing models developed on the historical data static settings. This model that was inspired by the principle of films uses “a sequence of snapshots, rather than a single photograph”. In this dynamic modeling framework, the customer credit risk is assessed using a batch data processing model. The model deals with the defaults in two ways. This Fig is about the full memory time window which is based on all previous data and the new data is appended to the training set. As you see at the Figure, learning 2 at month2 includes learning 1 which belongs to the months 1