US 12,243,067 B2
Machine learning based federated learning with hierarchical modeling hotel upsell
Andrew Vakhutinsky, Sharon, MA (US); Jorge Luis Rivero Perez, Naples, FL (US); Kirby Bosch, Ann Arbor, MI (US); and Recep Yusuf Bekci, Montreal (CA)
Assigned to Oracle International Corporation, Redwood Shores, CA (US)
Filed by Oracle International Corporation, Redwood Shores, CA (US)
Filed on Oct. 25, 2022, as Appl. No. 18/049,402.
Claims priority of provisional application 63/368,399, filed on Jul. 14, 2022.
Prior Publication US 2024/0020716 A1, Jan. 18, 2024
Int. Cl. G06Q 30/0201 (2023.01); G06N 7/01 (2023.01); G06N 20/20 (2019.01); G06Q 10/02 (2012.01); G06Q 50/12 (2012.01)
CPC G06Q 30/0206 (2013.01) [G06N 20/20 (2019.01); G06Q 10/02 (2013.01); G06Q 30/0201 (2013.01); G06Q 50/12 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method of upselling a hotel room selection, the method comprising:
generating a first hierarchical prediction model corresponding to a first hotel chain, the first hierarchical prediction model receiving reservation data from one or more corresponding first hotel properties, wherein the first hierarchical prediction model comprises a first Multinomial Logit (MNL) model using Bayesian inference;
generating a second hierarchical prediction model corresponding to a second hotel chain, the second hierarchical prediction model receiving reservation data from one or more corresponding second hotel properties, wherein the second hierarchical prediction model comprises a second Multinomial Logit (MNL) model using Bayesian inference;
generating the first hierarchical prediction model and generating the second hierarchical prediction model comprises receiving a plurality of textual room descriptions that define different types of hotel rooms, data mining the plurality of textual room descriptions to generate a plurality of features, and using at least a subset of the plurality of features to train the first MNL model and the second MNL model;
at each of the first hierarchical prediction model and the second hierarchical prediction model, generating corresponding model parameters, the model parameters comprising at least one of a feature value, a price sensitivity coefficient or a display position coefficient;
at a horizontal federated server, receiving the corresponding model parameters and averaging the model parameters to be used as a new probability distribution for the Bayesian inference;
distributing the new probability distribution to the first hotel properties and the second hotel properties; and
using the new probability distribution, retraining the first hierarchical prediction model and the second hierarchical prediction model.