Massachusetts Institute of Technology

Oracle International Corporation

Oracle International Corpration

Georgia Perakis

Kiran V Panchamgam

Leann Thayaparan

Setareh Borjian

Maxime Cohen

Jeremy Kalas

Embodiments generate and train an optimized demand model for predicting a demand of an item. Embodiments receive a plurality of trees, each of the plurality of trees including one or more levels of splits and a plurality of nodes, each of the plurality of nodes corresponding to a demand feature that influences demand for the item. Embodiments store 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, embodiments select a first demand feature that a greatest number of the plurality of trees split on. Embodiments optimize the first demand feature using the stored current bound to generate a second bound. Embodiments store the second bound as the current bound for each of the plurality of trees and move down each of the plurality of trees to a next level of splits.

Optimized tree ensemble based demand model

Embodiments generate an optimized demand model for a retail item. Embodiments train a tree ensemble machine learning model comprising a plurality of trees, the training comprising storing upper bounds for each of the trees, the trees comprising levels and branches that correspond to the demand features that influence demand for the item. Embodiments generate an objective function for the demand model. At a top split of each tree, embodiments determine optimal child nodes using the stored upper bounds and calculate a new feasible region for each tree. Using bounds on the new feasible region, embodiments move down each tree to a next level of splits and generate the optimized demand model when a leaf node of every tree has been reached.

Systems, methods, and other embodiments associated with determining a promotion price schedule for each item in a group are described. In one embodiment, a method includes computing an item coefficient that corresponds to a change in a value of an objective function when the item is priced at the promotion price. The objective function is based on a multiple product demand model. An item coefficient is computed for each item, each time period in the price schedule, and each promotion price in a price ladder for the item. An approximate objective function is formulated that includes products of item coefficients and binary decision variables. The item coefficients, the approximate objective function, and constraints are provided to an optimizer that determines values of the decision variables that maximize the approximate objective function. A promotion price schedule is created for each item based on values of the decision variables.

Growing community of inventors

Belmont, MA, United States of America

Georgia Perakis