Method and system for generating ensemble demand forecasts

Patent metadata

Field	Value
Publication number	US-12002063-B2
Application number	US-202217850640-A
Country	US
Kind code	B2
Filing date	Jun 27, 2022
Priority date	Oct 26, 2018
Publication date	Jun 4, 2024
Grant date	Jun 4, 2024

Abstract

Methods and systems for forecasting demand for a plurality of items are provided. In particular, the demand forecasting system and methods described herein are useful for predicting demand of products in a retail context. Forecast models are built and used to score incoming sales data to predict future demand for items. Forecast models are validated by evaluating actual demand against predicted demand and using that information to inform how future ensemble forecast will be generated. Forecasts may be broken down into smaller components to satisfy a variety of requests for data from client applications.

First claim

The invention claimed is:

1. A system for forecasting demand for a plurality of items, the system comprising:
a common data preparation computing subsystem including a standard data store, processor, and memory communicatively coupled to the processor, the memory storing instructions executable by the processor to:
receive sales data having different formats from a plurality of different sources;
incrementally reformat the sales data as it is received to common format data to enable use of the sales data by each component model of a plurality of component models; and
store the common format data in the standard data store;
an enterprise forecast computing subsystem including a processor, and a memory communicatively coupled to the processor, the memory storing instructions executable by the processor to:

build an ensemble forecasting model using at least two component models of the plurality of component models, wherein building the ensemble forecasting model comprises a selecting of the at least two or more component based on past performance data and a weighting of the at least two component models relative to each other based on past forecasting performance data;
access the common format data;
generate an aggregate demand forecast using the ensemble forecasting model and the common format data;
communicate the aggregate demand forecast to a forecast data store for storage;
access the standard data store to receive real-time updates to the common format data; and
in response to receiving the real-time updates to the common format data, update the ensemble forecasting model by modifying the weighting, relative to each other, of the at least two component models of the plurality of component models used to build the ensemble forecasting model; and
a server comprising an API accessible by one or more client applications, the API configured to:
receive requests for demand forecasts,
communicate demand forecast requests to the forecasts data store, and
receive demand forecasts from the forecasts data store.

2. The system of claim 1 , wherein the ensemble forecasting model is calculated using an affine function.

3. The system of claim 1 , wherein the plurality of component models includes one or more of an ARIMA model or a LOESS model.

4. The system of claim 1 , wherein the API is further configured to generate an administrator user interface for visualizing demand forecast data.

5. The system of claim 1 , further comprising a cloud platform with at least one load balancer to direct requests to available servers for processing.

6. The system of claim 1 , wherein the past forecasting performance data is evaluated by a forecast validation subsystem by running multiple forecasts for one set of sales data and cross-validating distributions of results.

7. A method of forecasting demand for a plurality of items within a retail enterprise, the method comprising:
receiving sales data at a common data preparation computing subsystem, the sales data being received in a plurality of different formats from a plurality of different sources;
incrementally reformatting, using the common data preparation computing subsystem, the sales data as it is received into common format data to enable use of the sales data by each component model of a plurality of component models;
storing the common format data in standard data store;
building, with an enterprise forecast computing subsystem, an ensemble forecast model by combining two or more component models of the plurality of component models, wherein building the ensemble forecast model comprises a selecting of the two or more component models and a weighting of the two or more component models relative to each other based on an assessment of past forecasting performance data;
automatically receiving real-time updates to the common format data from the standard data store;
updating the ensemble forecast model in response to receiving the real-time updates to the common format data by modifying the weighting, relative to each other, of the two or more component models of the plurality of component models used to build the ensemble forecast model;
analyzing, with the ensemble forecast model, the common format data to generate an aggregate demand forecast; and
storing the aggregate demand forecast in a forecasts data store;
receiving, at an API of a server accessible by one or more client applications, requests for demand forecasts;
communicate, from the API of the server, demand forecast request to the forecasts data store; and
receive, at the API of the server, demand forecasts from the forecasts data store.

8. The method of claim 7 , further comprising:
receiving, at an API, a request from a client application for a demand forecast;
querying the forecasts data store for the appropriate aggregate demand forecast; and
responding to the client application with the requested demand forecast.

9. The method of claim 7 , wherein the aggregate demand forecast is generated for an individual item, over a 1 week time period, for all sales of the retailer.

10. The method of claim 8 , wherein the request comprises a time period, a location, and one or more items.

11. The method of claim 8 , wherein the request is communicated to the data store by a real-time query service.

12. The method of claim 8 , wherein the aggregate demand forecast is visualized on a user interface for viewing and analysis by an administrator user.

13. The method of claim 8 , further comprising validating the ensemble forecast model by:
receiving a validation and selection of configuration options at a command line tool;
sending a submission packet to a validation server;
querying a demand forecast data store;
storing a validation set in the data repository;
calibrating the model with historical training data;
testing the model by calculating prediction for each set of forecast coordinates;
saving the predicted values;
calculating forecast validation results and store in data repository; and
displaying visualization of forecast performance on a validation user interface.

14. The method of claim 7 , wherein updating the ensemble forecast model comprises removing one of the two or more component models from the ensemble forecast model and selecting a different model from the plurality of component models to build the ensemble forecast model.

15. The method of claim 7 , wherein weighting the two or more component models based on the assessment of past forecasting performance data comprises weighting a first model of the two or more component models more heavily than a second model of the two or more component models based on a superior performance of the first model than the second model in the past forecasting performance data.

16. The method of claim 7 , wherein modifying the weighting of the two or more component models of the plurality of component models used to build the ensemble forecast model is based in part on forecasting performance data of the two or more component models for the updates to the common format data.

17. The method of claim 7 , wherein the updates to the common format data are based on a change in season or item promotions.

18. The method of claim 7 , wherein updating the ensemble forecast model is continually performed in response to real-time updates to the common format data.

19. A non-transitory computer-readable storage medium comprising computer-executable instruction which, when executed by a computing system, cause the computing system to perform a method of generating

Assignees

• Target Brands Inc

Inventors

• Morgan Adam
• Das Anand
• Shah Samir H
• Yelland Phillip

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06Q30/0202 Primary

  Market predictions or forecasting for commercial activities · CPC title

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title

• G06Q10/087 Primary

  Inventory or stock management, e.g. order filling, procurement or balancing against orders · CPC title