Method and system for performing data prediction

Patent metadata

Field	Value
Publication number	US-12566985-B2
Application number	US-202217687329-A
Country	US
Kind code	B2
Filing date	Mar 4, 2022
Priority date	Mar 4, 2022
Publication date	Mar 3, 2026
Grant date	Mar 3, 2026

Abstract

A method for performing data prediction includes: obtaining a dataset; generating prediction parameters using the dataset; identifying significant variables in the dataset; predicting seasonality of the dataset based on the significant variables; determining uncertainty of the prediction parameters; performing the data prediction by minimizing randomness and uncertainty of the dataset; and displaying the data prediction on a graphical user interface.

First claim

What is claimed is:

1 . A method for performing data prediction, the method comprising:
receiving a data prediction request from a user of a client device via a graphical user interface (GUI),
wherein the data prediction request comprises information identifying a dataset and information identifying a prediction feature in the dataset;
obtaining, in response to the data prediction request, the dataset,
wherein the dataset is related to a merchant,
wherein the prediction feature affects an event of a plurality of events demonstrated by the dataset, wherein the plurality of events comprises changes in accounts receivable;
determining, by a circuitry using an augmented dickey fuller test on the dataset, that the dataset has uncertainty; in response to determining that the dataset has uncertainty:
generating prediction parameters using the dataset, wherein the prediction parameters comprise region, sub-region, time-based accounts receivables of the merchant, and time-based accounts receivables difference of the merchant;
identifying significant variables in the dataset;
determining uncertainty of the prediction parameters using an ensemble of prediction models;
performing, based on the significant variables and the uncertainty of the prediction parameters, the data prediction using a combination of an extreme gradient boost (XGBoost) model, a multivariate adaptive regression splines (MARS) model, and a Bayesian model averaging (BMA) approach to minimize the uncertainty of the dataset; and
displaying the data prediction to the user using the GUI,
wherein the GUI is displayed on a display of a computing device,
wherein the display shows the data prediction in a user-friendly visual format that would allow the user to easily read and parse the data prediction, wherein the user-friendly visual format comprises subtabs corresponding to each data prediction of a plurality of data predictions, wherein the plurality of data predictions comprises the data prediction.

2 . The method of claim 1 , further comprising:
prior to determining the uncertainty of the prediction parameters, obtaining posterior probabilities of the ensemble of prediction models.

3 . The method of claim 2 , wherein the posterior probabilities of the ensemble of prediction models are obtained using the BMA approach.

4 . The method of claim 1 , wherein the significant variables in the dataset are identified using the MARS model.

5 . The method of claim 4 , wherein the seasonality of the dataset is predicted using the XGBoost model and the MARS model.

6 . The method of claim 1 ,
wherein the method further comprises:
obtaining posterior probabilities of an ensemble of prediction models,
wherein the significant variables in the dataset are identified using the MARS model; wherein the posterior probabilities of the ensemble of prediction models are obtained using the BMA approach; wherein the seasonality of the dataset is predicted using the XGBoost model and the MARS model; and wherein the data prediction is performed by an optimization module, wherein the optimization module comprises the MARS model, the BMA approach, and the XGBoost model.

7 . A non-transitory computer readable medium comprising computer readable program code, which when executed by a computer processor enables the computer processor to perform a method for performing data prediction, the method comprising:
receiving a data prediction request from a user of a client device via a graphical user interface (GUI),
wherein the data prediction request comprises information identifying a dataset and information identifying a prediction feature in the dataset;
obtaining, in response to the data prediction request, the dataset,
wherein the dataset is related to a merchant,
wherein the prediction feature affects an event of a plurality of events demonstrated by the dataset, wherein the plurality of events comprises changes in accounts receivable;
determining, by a circuitry using an augmented dickey fuller test on the dataset, that the dataset has uncertainty; in response to determining that the dataset has uncertainty:
generating prediction parameters using the dataset, wherein the prediction parameters comprise region, sub-region, time-based accounts receivables of the merchant, and time-based accounts receivables difference of the merchant;
identifying significant variables in the dataset;
determining uncertainty of the prediction parameters using an ensemble of prediction models;
performing, based on the significant variables and the uncertainty of the prediction parameters, the data prediction using a combination of an extreme gradient boost (XGBoost) model, a multivariate adaptive regression splines (MARS) model, and a Bayesian model averaging (BMA) approach to minimize the uncertainty of the dataset; and
displaying the data prediction to the user using the GUI,
wherein the GUI is displayed on a display of a computing device,
wherein the display shows the data prediction in a user-friendly visual format that would allow the user to easily read and parse the data prediction, wherein the user-friendly visual format comprises subtabs corresponding to each data prediction of a plurality of data predictions, wherein the plurality of data predictions comprises the data prediction.

8 . The non-transitory computer readable medium of claim 7 , further comprising:
prior to determining the uncertainty of the prediction parameters, obtaining posterior probabilities of the ensemble of prediction models.

9 . The non-transitory computer readable medium of claim 8 , wherein the posterior probabilities of the ensemble of prediction models are obtained using the BMA approach.

10 . The non-transitory computer readable medium of claim 7 , wherein the significant variables in the dataset are identified using the MARS model.

11 . The non-transitory computer readable medium of claim 10 , wherein the seasonality of the dataset is predicted using the XGBoost model and the MARS model.

12 . The non-transitory computer readable medium of claim 7 ,
wherein the method further comprises:
obtaining posterior probabilities of an ensemble of prediction models,
wherein the significant variables in the dataset are identified using the MARS model; wherein the posterior probabilities of the ensemble of prediction models are obtained using the BMA approach; wherein the seasonality of the dataset is predicted using the XGBoost model and the MARS model; and wherein the data prediction is performed by an optimization module, wherein the optimization module comprises the MARS model, the BMA approach, and the XGBoost model.

13 . A system for performing data prediction, the system comprising:
a processor comprising circuitry; memory comprising instructions, which when executed perform a method, the method comprising:
receiving a data prediction request from a user of a client device via a graphical user interface (GUI),
wherein the data prediction request comprises information identifying a dataset and information identifying a prediction feature in the dataset;

obtaining, in response to the data prediction request, the dataset,
wherein the dataset is related to a merchant,
wherein the prediction feature affects an event of a plurality of events demonstrated by the dataset, wherein the plurality of events comprises changes in accounts receivable;

determining, by the circuitry using an augmented dickey fuller test on the dataset, that the dataset has uncertainty;
in response to determining that the dataset has uncertainty:
generating prediction parameters using the dataset,

Assignees

• Dell Products Lp

Inventors

• Cv Raghavendra Rao
• Dan Rajkumar
• Ganguly Priyanka
• S Dilip Kumar
• Ramakrishnan Hariraj

Classifications

• G06N20/20

  Ensemble learning · CPC title

• G06N7/01 Primary

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