HEAD-DT evolutionary scheme. Figure 2 presents the evolutionary scheme followed by HEAD-DT. A random initial population of individuals (decision-tree algorithms) is created and evaluated according to the performance of their corresponding trees in a meta-training set. Then, a selection procedure is responsible for choosing individuals that will undergo breeding operations. After a new population is complete, it is once again evaluated and the process continues until a maximum number of generations is reached. The best decision-tree induction algorithm is then executed over a meta-test set, which estimates its performance in unseen data.
Barros et al. BMC Bioinformatics 2012 13:310 doi:10.1186/1471-2105-13-310