US 12,307,404 B2
Optimized tree ensemble based demand model
Leann Thayaparan, Manlius, NY (US); Kiran V. Panchamgam, Bedford, MA (US); Setareh Borjian, New Haven, CT (US); and Georgia Perakis, Belmont, MA (US)
Assigned to Oracle International Corporation, Redwood Shores, CA (US); and MASSACHUSETTS INSITUTE OF TECHNOLOGY, Cambridge, MA (US)
Filed by Oracle International Corporation, Redwood Shores, CA (US); and MASSACHUSETTS INSTITUTE OF TECHNOLOGY, Cambridge, MA (US)
Filed on May 16, 2024, as Appl. No. 18/666,037.
Application 18/666,037 is a continuation of application No. 17/449,112, filed on Sep. 28, 2021, granted, now 12,020,196.
Prior Publication US 2024/0303577 A1, Sep. 12, 2024
This patent is subject to a terminal disclaimer.
Int. Cl. G06Q 10/067 (2023.01); G06F 18/214 (2023.01); G06N 20/20 (2019.01); G06Q 10/08 (2023.01); G06Q 10/087 (2023.01)
CPC G06Q 10/067 (2013.01) [G06F 18/2148 (2023.01); G06N 20/20 (2019.01); G06Q 10/08 (2013.01); G06Q 10/087 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method implemented by one or more processors for generating and training an optimized demand model for predicting a demand of an item, the method of generating and training the optimized demand model comprising:
training a tree ensemble machine learning model comprising a plurality of trees by:
receiving the plurality of trees, each of the plurality of trees comprising one or more levels of splits and a plurality of nodes, each of the plurality of nodes corresponding to a demand feature that influences the demand for the item;
storing a first bound as a current bound for each of the plurality of trees;
starting at a top split of each of the plurality of trees, selecting a first demand feature that a greatest number of the plurality of trees split on;
optimizing the first demand feature using the current bound to generate a second bound;
storing the second bound as the current bound for each of the plurality of trees;
moving down each of the plurality of trees to a next level of splits; and
at the next level of splits, for each tree, repeating the selecting the first demand feature that the greatest number of the plurality of trees split on, the optimizing the first demand feature using the current bound to generate the second bound, storing the second bound as the current bound for each of the plurality of trees and moving down each of the plurality of trees to the next level of splits until a leaf node of each of the plurality of trees has been reached; and
generating the optimized demand model when the leaf node of each of the plurality of trees has been reached using the current bound for each of the plurality of trees.