Installation
About
Supervised, multivariate, and non-parametric discretization algorithm based on tree ensembles learning and moment matching optimization. This version of the algorithm relies on random forest algorithm to learn a large set of split points that conserves the relationship between attributes and the target class, and on moment matching optimization to transform this set into a reduced number of cut points matching as well as possible statistical properties of the initial set of split points. For each attribute to be discretized, the set S of its related split points extracted through random forest is mapped to a reduced set C of cut points of size k. This mapping relies on minimizing, for each continuous attribute to be discretized, the distance between the four first moments of S and the four first moments of C subject to some constraints. This non-linear optimization problem is performed using k values ranging from 2 to 'max_splits', and the best solution returned correspond to the value k which optimum solution is the lowest one over the different realizations. ForestDisc is a generalization of RFDisc discretization method initially proposed by Berrado and Runger (2009) doi:10.1109/AICCSA.2009.5069327, and improved by Berrado et al. in 2012 by adopting the idea of moment matching optimization related by Hoyland and Wallace (2001) doi:10.1287/mnsc.47.2.295.9834.
Key Metrics
Downloads
Yesterday | 10 0% |
Last 7 days | 44 -17% |
Last 30 days | 152 +7% |
Last 90 days | 395 -38% |
Last 365 days | 1.884 -19% |