Granted Patent B2

US 12,591,908 · App. 18/749,896 · Granted Mar 31, 2026

Inventors: Kanchana Padmanabhan (Ottawa, CA); Anneya Golob (Ottawa, CA); Brian Keng (Ottawa, CA)

Assignee: Kinaxis Inc.

G06Q30/0244G06F40/40G06N5/04G06N20/00G06Q30/0201G06Q30/0202G06Q30/0264G06Q30/0276

Abstract

Claims (63)

1 . A system comprising:

a processor; and

a memory storing instructions that, when executed by the processor, configure the system to:

receive, by an AI demand forecasting engine, historical data related to sales of items and promotion design information;

transform, by a feature engineering pipeline, the promotion design information into numeric features for use in training predictions for item demand;

train, by the AI demand forecasting engine to predict independent baseline demands and independent promotion demands, based on the historical data and the numeric features;

generate, by the AI demand forecasting engine, a plurality of entities, each entity defined by a placement of a sale item in a promotional platform;

forecast, by the AI demand forecasting engine, a baseline forecast associated with each entity and a promotional forecast associated with each entity;

forecast, by the AI demand forecasting engine, an objective associated with each entity based on the baseline forecast and the promotional forecast;

generate by an optimization engine encoded with a mix of real integers and Boolean conditions, a plurality of plans, each plan consisting of a unique subset of entities;

generate, by the optimization engine, one or more candidate solutions based on each of the plurality of plans and by eliminating plans that violate at least one constraint encoded as at least one Boolean variable of the Boolean conditions; and

select, by the optimization engine, an optimum plan from the one or more candidate solutions based on optimization of the objective.

2 . The system of claim 1 , wherein the promotion design information comprises: flyer design information; television advertise information; store type; store region; loyalty programs; and promotion mechanics.

3 . The system of claim 1 , wherein the historical data is related to: holidays and seasonality; ring and regular price; trends and novelty; and item attributes.

4 . The system of claim 1 , wherein:

generating the one or more candidate solutions is based on evaluating a linear equation for each of the plurality of plans;

and wherein:

the linear equation is based on: the plurality of entities; the objective associated with each entity; and the at least one constraint;

and wherein:

the plurality of entities are each represented by a first Boolean value, and

each constraint is represented by a second Boolean value.

5 . The system of claim 1 , wherein the AI demand forecast engine uses a machine learning model selected from at least one of: Deep Learning, Gradient Boosted Trees, Random Forest, Kernel Density Estimators, Gaussian Processes, Isolation Forests, Generalized Additive Models, Representation Learning, Non-parametric techniques, Econometric Models, Bayesian Models, Time-Series Models, and Bayesian Additive Regression Trees.

6 . A non-transitory computer-readable storage medium, the computer-readable storage medium including instructions that when executed by a computer, cause the computer to:

receive, by an AI demand forecasting engine, historical data related to sales of items and promotion design information;

transform, by a feature engineering pipeline, the promotion design information into numeric features for use in training predictions for item demand;

train, by the AI demand forecasting engine to predict independent baseline demands and independent promotion demands, based on the historical data and the numeric features;

generate, by the AI demand forecasting engine, a plurality of entities, each entity defined by a placement of a sale item in a promotional platform;

forecast, by the AI demand forecasting engine, a baseline forecast associated with each entity and a promotional forecast associated with each entity;

forecast, by the AI demand forecasting engine, an objective associated with each entity based on the baseline forecast and the promotional forecast;

generate by an optimization engine encoded with a mix of real integers and Boolean conditions, a plurality of plans, each plan consisting of a unique subset of entities;

generate, by the optimization engine, one or more candidate solutions based on each of the plurality of plans and by eliminating plans that violate at least one constraint encoded as at least one Boolean variable of the Boolean conditions; and

select, by the optimization engine, an optimum plan from the one or more candidate solutions based on optimization of the objective.

7 . The computer-readable storage medium of claim 6 , wherein the promotion design information comprises: flyer design information;

television advertise information; store type; store region; loyalty programs; and promotion mechanics.

8 . The computer-readable storage medium of claim 6 , wherein the historical data is related to: holidays and seasonality; ring and regular price; trends and novelty; and item attributes.

9 . The computer-readable storage medium of claim 6 , wherein:

generating the one or more candidate solutions is based on evaluating a linear equation for each of the plurality of plans;

and wherein:

the linear equation is based on: the plurality of entities; the objective associated with each entity; and the at least one constraint;

and wherein:

the plurality of entities are each represented by a first Boolean value, and

each constraint is represented by a second Boolean value.

10 . The computer-readable storage medium of claim 6 , wherein the AI demand forecast engine uses a machine learning model selected from at least one of: Deep Learning, Gradient Boosted Trees, Random Forest, Kernel Density Estimators, Gaussian Processes, Isolation Forests, Generalized Additive Models, Representation Learning, Non-parametric techniques, Econometric Models, Bayesian Models, Time-Series Models, and Bayesian Additive Regression Trees.

11 . A computer-implemented method comprising:

receiving, by an AI demand forecasting engine, historical data related to sales of items and promotion design information;

transforming, by a feature engineering pipeline, the promotion design information into numeric features for use in training predictions for item demand;

training, by the AI demand forecasting engine to predict independent baseline demands and independent promotion demands, based on the historical data and the numeric features;

generating, by the AI demand forecasting engine, a plurality of entities, each entity defined by a placement of a sale item in a promotional platform;

forecasting, by the AI demand forecasting engine, a baseline forecast associated with each entity and a promotional forecast associated with each entity;

forecasting, by the AI demand forecasting engine, an objective associated with each entity based on the baseline forecast and the promotional forecast;

generating by an optimization engine encoded with a mix of real integers and Boolean conditions, a plurality of plans, each plan consisting of a unique subset of entities;

generating, by the optimization engine, one or more candidate solutions based on each of the plurality of plans and by eliminating plans that violate at least one constraint encoded as at least one Boolean variable of the Boolean conditions; and

selecting, by the optimization engine, an optimum plan from the one or more candidate solutions based on optimization of the objective.

12 . The computer-implemented method of claim 11 , wherein the promotion design information comprises: flyer design information; television advertising information; store type; store region; loyalty programs; and promotion mechanics.

13 . The computer-implemented method of claim 11 , wherein the historical data is related to: holidays and seasonality; ranging and regular price; trends and novelty; and item attributes.

14 . The computer-implemented method of claim 11 , wherein:

generating the one or more candidate solutions is based on evaluating a linear equation for each of the plurality of plans;

and wherein:

the linear equation is based on: the plurality of entities; the objective associated with each entity; and the at least one constraint;

and wherein:

the plurality of entities are each represented by a first Boolean value, and

each constraint is represented by a second Boolean value.

15 . The computer-implemented method of claim 11 , wherein the AI demand forecasting engine uses a machine learning model selected from at least one of: Deep Learning, Gradient Boosted Trees, Random Forest, Kernel Density Estimators, Gaussian Processes, Isolation Forests, Generalized Additive Models, Representation Learning, Non-parametric techniques, Econometric Models, Bayesian Models, Time-Series Models, and Bayesian Additive Regression Trees.

Assignments (1)

Continuity (3)

References Cited (29)