Selecting optimum primary and secondary parameters to calibrate and generate an unbiased forecasting model

Patent metadata

Field	Value
Publication number	US-12020124-B2
Application number	US-202016743952-A
Country	US
Kind code	B2
Filing date	Jan 15, 2020
Priority date	Jan 15, 2020
Publication date	Jun 25, 2024
Grant date	Jun 25, 2024

Abstract

A device may receive demand data associated with a product or a service, multiple forecasting models, and multiple cost functions, and may identify primary parameters for the multiple models based on the demand data. The device may utilize a model to rank the multiple forecasting models and the multiple cost functions based on the primary parameters, and may select optimum primary parameters based on ranking the multiple forecasting models and the multiple cost functions. The device may identify secondary parameters for the multiple forecasting models based on the demand data and the optimum primary parameters. The device may select optimum secondary parameters based on ranking the multiple forecasting models and the multiple cost functions, and may select a forecasting model, from the multiple forecasting models, based on the optimum primary parameters and the optimum secondary parameters. The device may perform one or more actions based on selecting the forecasting model.

First claim

What is claimed is:

1. A method, comprising:
receiving, by a device, demand data, a plurality of forecasting models, and a plurality of cost functions;
training, by the device, a machine learning model;
processing, by the device, the plurality of forecasting models, with the machine learning model, to generate a plurality of optimized forecasting models;
identifying, by the device, primary parameters for the plurality of optimized forecasting models based on the demand data,
wherein the primary parameters include one or more of:
an exhaustive true/false search for all variables,
a false model average of variables per a Bayesian framework, or
reconsidered true/false variables;

utilizing, by the device, a model to rank the plurality of optimized forecasting models and the plurality of cost functions based on the primary parameters,
wherein utilizing the model to rank the plurality of optimized forecasting models and the plurality of cost functions comprises:
determining a mean absolute percentage error to measure a prediction accuracy of each of the plurality of optimized forecasting models and each of the plurality of cost functions based on the primary parameters;

selecting, by the device, a set of the primary parameters based on ranking the plurality of optimized forecasting models and the plurality of cost functions,
wherein the set of the primary parameters are optimum for the plurality of optimized forecasting models as compared to the primary parameters not included in the set of the primary parameters;

identifying, by the device, secondary parameters for the plurality of forecasting models based on the demand data and the set of the primary parameters;
selecting, by the device, a set of the secondary parameters based on ranking the plurality of optimized forecasting models and the plurality of cost functions,
wherein the set of the secondary parameters are optimum for the plurality of optimized forecasting models as compared to the secondary parameters not included in the set of the secondary parameters;

selecting, by the device, a forecasting model, from the plurality of optimized forecasting models, to generate a calibrated and unbiased forecasting model that utilizes the primary parameters and the secondary parameters to conserve computing resources based on the set of the primary parameters and the set of the secondary parameters; and
performing, by the device, one or more actions based on selecting the forecasting model,
where performing the one or more actions comprises:
providing updated data associated with the plurality of forecasting models to retrain the machine learning model in order to update the machine learning model; and
implementing pricing actions for one or more of:
coupons,
rebates,
wholesale prices, or
sale prices.

2. The method of claim 1 , wherein the plurality of optimized forecasting models include a plurality of time series decomposition models.

3. The method of claim 1 , wherein selecting the set of the primary parameters comprises:
processing the demand data, with the plurality of optimized forecasting models and based on the primary parameters and ranking the plurality of forecasting models, to generate output data;
processing the demand data and the output data, with the plurality of cost functions and based on ranking the plurality of cost functions, to determine differences between the demand data and the output data; and
selecting the set of the primary parameters based on the differences between the demand data and the output data.

4. The method of claim 1 , wherein selecting the set of the secondary parameters comprises:
processing the demand data, with the plurality of optimized forecasting models and based on the secondary parameters and ranking the plurality of optimized forecasting models, to generate output data;
processing the demand data and the output data, with the plurality of cost functions and based on ranking the plurality of cost functions, to determine differences between the demand data and the output data; and
selecting the set of the secondary parameters based on the differences between the demand data and the output data.

5. The method of claim 1 , wherein performing the one or more actions comprises one or more of:
providing the forecasting model to a client device or a server device;
utilizing the forecasting model to implement one or more pricing actions for a product or service;
utilizing the forecasting model to implement one or more promotions for a product or service;
calibrating the forecasting model based on a Bayesian tuning technique; or
generating one or more predictions with the forecasting model.

6. The method of claim 1 , wherein selecting the set of the primary parameters comprises:
simultaneously processing the demand data, with each of the plurality of optimized forecasting models and based on a rank of each of the plurality of forecasting models and the primary parameters, to generate output data;
simultaneously processing the demand data and the output data, with each of the plurality of cost functions and based on a rank of each of the plurality of cost functions, to determine differences between the demand data and the output data; and
selecting the set of the primary parameters based on the differences between the demand data and the output data.

7. A device, comprising:
one or more memories; and
one or more processors, communicatively coupled to the one or more memories, configured to:
receive demand data associated a service, a plurality of forecasting models, and a plurality of cost functions;
train a machine learning model;
process the plurality of forecasting models, with the machine learning model, to generate a plurality of optimized forecasting models;
identify primary parameters for the plurality of forecasting models based on the demand data,
wherein the primary parameters include one or more of:
an exhaustive true/false search for all variables, or
a false model average of variables per a Bayesian framework;

utilize a model to rank the plurality of forecasting models and the plurality of cost functions based on the primary parameters,
wherein the one or more processors, to utilize the model to rank the plurality of optimized forecasting models and the plurality of cost functions, are to:
determine a mean absolute percentage error to measure a prediction accuracy of each of the plurality of optimized forecasting models and each of the plurality of cost functions based on the primary parameters;

select a set of the primary parameters based on ranking the plurality of forecasting models and the plurality of cost functions,
wherein the set of the primary parameters are optimum for the plurality of forecasting models as compared to the primary parameters not included in the set of the primary parameters;

identify secondary parameters for the plurality of forecasting models based on the demand data and the set of the primary parameters;
select a set of the secondary parameters based on ranking the plurality of forecasting models and the plurality of cost functions,
wherein the set of the secondary parameters are optimum for the plurality of forecasting models as compared to the secondary parameters not included in the set of the secondary parameters;

select a forecasting model, from the plurality of forecasting models, to generate a calibrated and unbiased forecasting model that utilizes the primary parameters and the secondary parameters to conserve computing resources based on the set of the primary parameters and the set of the secondary parameters; and
perform one or more actions based on selecting the forecasting model,
wherein the one or more processors, to perform the one or more action

Assignees

• Accenture Global Solutions Ltd

Inventors

• Husain Afzal
• Aswathnarayana Suresh
• Parikh Ashesh
• Astro Paolo
• Monroy Mercedes

Classifications

• G06Q10/0633

  Workflow analysis · CPC title

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06Q30/0202 Primary

  Market predictions or forecasting for commercial activities · CPC title

• G06N20/00 Primary

  Machine learning · CPC title