Figure 6.

The flowchart of the PreDBA method. (A) The original training data (D) has m complex individuals and n characteristics (so it is m x n). (B) There are three various machine learning models (GBRT, AdaR, BagR) before training on D. (C) All model gives a forecast of the event, which is then cast into the second level of training data (Dnew), which is now m × 3. That is, three forecasts become characteristics of the second layer machine learning model (s). (D) The characteristics are trained on the second level model (XGBR) to produce the ultimate outcome.