Tous les livres

Professionnel

Applis À propos Coach Connexion Commencer la lecture

Quantitative Finance · Glossaire

Qu'est-ce que « Monotonic constraint » ?

Definition 5.7 Machine Learning for Markets · Chapitre 5 — Trees and Boosting

A monotonic constraint restricts a model to be non-decreasing (or non-increasing) in a feature, all others fixed. In a tree ensemble it is enforced by allowing a split on that feature only when the left child’s value does not exceed the right child’s (or the reverse), propagated down the tree.

Partial dependence of the five-second forecast on the order-flow imbalance (the feature swept over its 2nd to 98th percentile, the others at their observed values), for boosted trees without and with monotonic constraints. Data: ml_trees.partial_dependence on firm.tape.
Figure 5.3. Partial dependence of the five-second forecast on the order-flow imbalance (the feature swept over its 2nd to 98th percentile, the others at their observed values), for boosted trees without and with monotonic constraints. Data: ml_trees.partial_dependence on firm.tape.
Lire dans le chapitre →