	US 12,443,909 B2
	System for modeling the performance of fulfilment machines
Lei Cao, Worcester, MA (US); Brian Leo Quanz, Yorktown Heights, NY (US); Ajay Ashok Deshpande, White Plains, NY (US); Xuan Liu, Yorktown Heights, NY (US); Arun Hampapur, Norwalk, CT (US); Ali Koc, White Plains, NY (US); Yingjie Li, Chappaqua, NY (US); and Yada Zhu, Westchester, NY (US)
Assigned to International Business Machines Corporation, Armonk, NY (US)
Filed by INTERNATIONAL BUSINESS MACHINES CORPORATION, Armonk, NY (US)
Filed on Jun. 11, 2018, as Appl. No. 16/005,081.
Prior Publication US 2019/0378061 A1, Dec. 12, 2019
Int. Cl. G06Q 10/067 (2023.01); G06F 30/20 (2020.01); G06N 3/08 (2023.01); G06Q 10/0631 (2023.01); G06Q 10/0637 (2023.01); G06Q 10/0639 (2023.01)

CPC G06Q 10/067 (2013.01) [G06F 30/20 (2020.01); G06N 3/08 (2013.01); G06Q 10/06315 (2013.01); G06Q 10/06375 (2013.01); G06Q 10/06393 (2013.01)]

18 Claims

1. A computing device comprising:

a processor;

a network interface coupled to the processor to enable communication over a network;

a storage device having a memory stack configured to store raw data and for content and programming, coupled to the processor; and

a fulfilment strategy program stored in the storage device, wherein an execution of the program by the processor configures the computing device to perform acts comprising:

receiving input parameters from a computing device of a user that is stored in the memory stack;

receiving historical data related to a network of nodes from a data repository that is stored in the memory stack as part of the raw data;

reducing a computational load and increasing an accuracy of the processor by soliciting constraints from the computing device of the user that reduce a volume of the historical data stored in the memory stack, to samples that are deemed to be relevant for calculations processed by the processor;

determining a synthetic demand status based on the reduced volume of historical data and the input parameters;

determining a synthetic network status based on the reduced volume of historical data and the input parameters;

identifying a fulfilment strategy based on the synthetic demand status and the synthetic network status;

increasing the accuracy in a machine learning model to determine how one or more factors influence the identified fulfilment strategy, by training the machine learning model by not only using the historical data, but also the synthetic demand status and the synthetic network status;

adjusting the machine learning model to achieve a least computationally complex model on the processor that meets a predetermined threshold level of accuracy, wherein the adjusting includes a least absolute shrinkage and selection operator (lasso) to automatically analyze a correlation between fulfilment parameters of the machine learning model by performing both variable selection and regularization in order to enhance a prediction accuracy and interpretability of the resulting adjusted machine learning model; and

determining key performance indicators (KPIs) for the fulfilment strategy based on the synthetic demand status and the synthetic network status, wherein:

the historical data includes data describing one or more orders; and

determining the synthetic demand status comprises, for each order in the historical data:

categorizing the order into one or more predetermined order categories;

identifying one or more order categories that most closely coincide with one or more input parameters of the one or more input parameters that relate to demand data; and

generating synthetic demand data based on the identified one or more order categories and the one or more input parameters that relate to demand data.