# US 2021/0241036 A1 · Computer-based systems including machine learning models trained on distinct dataset types and methods of use thereof

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0241036 A1
- Document ID: US20210241036A1
- Status: Published application
- Kind: A1
- Application number: 16/781,718
- Filed: 2020-02-04
- Published: 2021-08-05
- Earliest claimed: 2020-02-04
- Assignees: Capital One Services, LLC
- Applicants: Capital One Services, LLC
- Inventors: Abdelkader M'Hamed Benkreira; Michael Mossoba; Joshua Edwards
- Web page: https://patentgrep.com/patent/US20210241036A1

## CPC

- G06Q 30/0206: {Price or cost determination based on market factors}
- G06F 16/951: Indexing; Web crawling techniques
- G06F 16/9537: Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
- G06F 18/2148: {characterised by the process organisation or structure, e.g. boosting cascade}
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/084: Backpropagation, e.g. using gradient descent
- G06N 3/09: Supervised learning
- G06N 20/00: Machine learning
- G06Q 30/0643: {graphically representing goods, e.g. 3D product representation}
- G06N 3/045: Combinations of networks
- G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G06N 20/10: using kernel methods, e.g. support vector machines [SVM]
- G06N 20/20: Ensemble learning
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

In order to facilitate machine learning for prediction using distinct dataset types, systems and methods include collecting content information from archived websites databases. Collecting historical event information from online sources, where the historical event information is associated with a plurality of historical events. Generating event-dependent products training datasets based on the content information and the historical event information, where the event-dependent content training datasets defines for content historical events that are associated with attributes of the content, attribute change of the content, or both. Training an attribute prediction machine learning model based on the event-dependent content training datasets. Applying the trained attribute prediction machine learning model to additional event information to predict, for content, a future attribute estimate, a future attribute change estimate, or both. Causing to display an indication representative of the future attribute estimate, the future attribute change estimate, or both.

## Claims

### Claim 1

A method comprising:
collecting, by at least one processor, product information for a plurality of products from one or more electronic archived websites record databases;
wherein the product information for each product of the plurality of products comprises:
i) one or more attributes, and
ii) a date associated with each attribute of the one or more attributes;
collecting, by the at least one processor, historical event information from one or more online sources;
wherein the historical event information is associated with a plurality of historical events;
wherein the historical event information comprises a date associated with each event of the plurality of historical events;
determining, by the at least one processor, at least one impacted product of the plurality of products based at least in part on the one or more attributes having an impact comprising a fluctuation exceeding a predetermined impact fluctuation value over a duration exceeding a predetermined impact fluctuation duration;
generating, by the at least one processor, at least one event-dependent products training dataset based at least in part on the product information and the historical event information;
wherein the at least one event-dependent products training dataset comprises the at least one impacted product of the plurality of products and at least one event of the plurality of historical events that is associated with at least one date of the impact;
determining, by the at least one processor, an impact period for each event of the plurality of historical events based on a duration of the impact following the date of each event;
training, by the at least one processor, an attribute prediction machine learning model based at least in part on the at least one event-dependent products training dataset to obtain a trained attribute prediction machine learning model;
wherein the trained attribute prediction machine learning model comprises model parameters trained to correlate each event of the plurality of historical events to:
i) the impact on each product of the plurality of products, and
ii) the impact period of each product of the plurality of products;
collecting, by the at least one processor, additional event information from the one or more online sources;
wherein the additional event information is associated with one or more additional events;
wherein the additional event information comprises a date associated with each event of the one or more additional events;
applying, by the at least one processor, the trained attribute prediction machine learning model to the additional event information to predict, for at least one product of the plurality of products, an additional impact and an additional impact period associated with the impact for each additional event of the at least one additional event;
wherein the additional impact comprises:
(a) a future attribute estimate,
(b) a future attribute change estimate, or
(c) both (a) and (b); and
causing to display, by the at least one processor, at least one indication representative of (a) the future attribute estimate, (b) the future attribute change estimate, or (c) both (a) and (b), on a screen of at least one computing device.

### Claim 2 (depends on 1)

The method of claim 1, further comprising:
determining, by the at least one processor, a respective attribute impact period of one or more attribute impact periods associated with each respective additional event based on a respective variation in a respective one or more attribute of each product following the respective additional event information;
wherein the attribute impact period comprises a set of dates for which the additional event information is correlated with each respective variation; and
determining, by the at least one processor, an impacted product dataset comprising each respective product of the plurality of products associated with each respective variation in the respective one or more attribute.

### Claim 3 (depends on 2)

The method of claim 2, further comprising:
generating, by the at least one processor, a first training dataset comprising at least one subset of selected products;
generating, by the at least one processor, a second training dataset comprising the impacted product dataset; and
generating, by the at least one processor, a third training dataset comprising post-impact product information;
wherein the post-impact product information comprises respective product information of respective selected products of the subset of selected products having a respective date after each respective attribute impact period; and
wherein training the at least one attribute prediction machine learning model comprises training a first attribute prediction machine learning model with the first training dataset, a second attribute prediction machine learning model with the second training dataset, and a third attribute prediction machine learning model with the third training dataset.

### Claim 4 (depends on 2)

The method of claim 2, further comprising:
determining, by the at least one processor, that at least one attribute impact period of one or more attribute impact periods has not ended;
generating, by the at least one processor, a first training dataset comprising at least one subset of selected products;
generating, by the at least one processor, a second training dataset comprising non-impacted product information;
wherein the non-impacted product information comprises respective product information of respective selected products of the subset of selected products having a respective attribute-posted date not during the at least one attribute impact period; and
wherein training the at least one attribute prediction machine learning model comprises training a first attribute prediction machine learning model with the first training dataset, and a second attribute prediction machine learning model with the second training dataset.

### Claim 5 (depends on 1)

The method of claim 1, further comprising receiving, by the at least one processor, a prediction request from the at least one computing device associated with a user; and
wherein the prediction request comprising the at least one particular product.

### Claim 6 (depends on 5)

The method of claim 5, wherein the prediction request is generated by a user selection using an attribute prediction graphical user interface at the at least one computing device associated with the user.

### Claim 7 (depends on 1)

The method of claim 1, wherein the product information further comprises a product category associated with each of the plurality of products.

### Claim 8 (depends on 7)

The method of claim 7, further comprising:
causing, by the at least one processor, to display a product research graphical user interface (GUI) on a screen of the at least one computing device;
wherein the product research GUI is configured to allow the at least one user to define a category view for at least one of the product category associated with each of the plurality of products; and
wherein the category view presents a purchase recommendation based on the at least one predicted future attribute, at least one predicted future attribute change, or both, for each of the plurality of products of each product category.

### Claim 9 (depends on 1)

The method of claim 1, further comprising:
detecting, by the at least one processor, a viewed product information displayed on a screen of the at least one computing device associated with the user;
receiving, by the at least one processor via an application programming interface (API), an automatic attribute prediction request for the viewed product information;
determining, by the at least one processor, the at least one particular product based on the viewed product information;
determining, by the at least one processor, a cheapest date prediction based on the at least one particular product; and
wherein the cheapest date prediction is based on the at least one predicted future attribute, at least one predicted future attribute change, or both predicted by the at least one attribute prediction machine learning model for the at least one particular product.

### Claim 10 (depends on 9)

The method of claim 9, wherein the cheapest date prediction comprises a date within one year into the future.

### Claim 11

A method, comprising:
collecting, by at least one processor, product information for a plurality of products from one or more electronic archived websites record databases;
wherein the product information for each product of the plurality of products comprises:
i) one or more attributes, and
ii) a date associated with each attribute of the one or more attributes;
collecting, the at least one processor, event information from the plurality of online sources;
wherein the event information is associated with a plurality of events;
wherein the event information comprises at least one respective date of each respective event;
applying, by the at least one processor, at least one trained attribute prediction machine learning model to the product information and the event information to predict, for at least one product of the plurality of products, an impact and an impact period associated with the impact for each event of the plurality of events;
wherein the additional impact comprises:
(a) a future attribute estimate,
(b) a future attribute change estimate, or
(c) both (a) and (b);
wherein the impact period for each event of the plurality of events based on a duration of the impact following the at least one respective date of each event
wherein an attribute prediction machine learning model is trained based at least in part on at least one event-dependent products training dataset to generate the at least one trained attribute prediction machine learning model;
wherein the at least one event-dependent products training dataset comprises:
i) at least one impacted product of the plurality of products based at least in part on the one or more attributes having an impact comprising a fluctuation exceeding a predetermined impact fluctuation value over a duration exceeding a predetermined impact fluctuation duration, and
ii) at least one event of the plurality of historical events that is associated with at least one date of the impact;
wherein the at least one event-dependent products training dataset defines for each impacted product of the at least one impacted product
at least one attribute impact period associated with the at least one event;
wherein the at least one attribute impact period identifies a respective time period over which the fluctuation associated with the at least one event has occurred;
causing to display, by the at least one processor, at least one indication representative of the at least one predicted future attribute, at least one predicted future attribute change, or both, on a screen of at least one computing device associated with a user.

### Claim 12 (depends on 11)

The method of claim 11, wherein the at least one trained attribute prediction machine learning model comprises:
a first trained attribute prediction machine learning model trained using a first training dataset comprising the at least one subset of selected products;
a second trained attribute prediction machine learning model trained using a second training dataset comprising impacted product information;
wherein the impacted product information comprises respective product information of respective selected products of the subset of selected products having a respective date during the at least one attribute impact period; and
a third trained attribute prediction machine learning model trained using a third training dataset comprising post-impact product information;
wherein the post-impact product information comprises respective product information of respective selected products of the subset of selected products having a respective date after the at least one attribute impact period.

### Claim 13 (depends on 11)

The method of claim 11, wherein the at least one trained attribute prediction machine learning model comprises:
a first trained attribute prediction machine learning model trained using a first training dataset comprising the at least one subset of selected products; and
a second trained attribute prediction machine learning model trained using a second training dataset comprising non-impacted product information;
wherein the non-impacted product information comprises respective product information of respective selected products of the subset of selected products having a respective attribute-posted date not during the at least one attribute impact period.

### Claim 14 (depends on 11)

The method of claim 11, further comprising receiving, by the at least one processor, a prediction request from the at least one computing device associated with the user; and
wherein the prediction request comprising the at least one particular product.

### Claim 15 (depends on 14)

The method of claim 14, wherein the prediction request is generated by a user selection using an attribute prediction graphical user interface at the at least one computing device associated with the user.

### Claim 16 (depends on 11)

The method of claim 11, wherein the product information further comprises a product category associated with each of the plurality of products.

### Claim 17 (depends on 16)

The method of claim 16, further comprising:
causing, by the at least one processor, to display a product research graphical user interface (GUI) on a screen of the at least one computing device;
wherein the product research GUI is configured to allow the at least one user to define a category view for at least one of the product category associated with each of the plurality of products; and
wherein the category view presents a purchase recommendation based on the at least one predicted future attribute, at least one predicted future attribute change, or both, for each of the plurality of products of each product category.

### Claim 18 (depends on 17)

The method of claim 17, further comprising:
detecting, by the at least one processor, a viewed product information displayed on a screen of the at least one computing device associated with the user;
receiving, by the at least one processor via an application programming interface (API), an automatic attribute prediction request for the viewed product information;
determining, by the at least one processor, the at least one particular product based on the viewed product information;
determining, by the at least one processor, a cheapest date prediction based on the at least one particular product; and
wherein the cheapest date prediction is based on the at least one predicted future attribute, at least one predicted future attribute change, or both predicted by the at least one attribute prediction machine learning model for the at least one particular product.

### Claim 19 (depends on 18)

The method of claim 18, wherein the cheapest date prediction comprises a date within one year into the future.

### Claim 20

A system, comprising:
at least one processor configured to:
collect product information for a plurality of products from one or more electronic archived websites record databases;
wherein the product information for each product of the plurality of products comprises:
i) one or more attributes, and
ii) a date associated with each attribute of the one or more attributes;
collect historical event information from one or more online sources;
wherein the historical event information is associated with a plurality of historical events;
wherein the historical event information comprises a date associated with each event of the plurality of historical events;
determine at least one impacted product of the plurality of products based at least in part on the one or more attributes having an impact comprising a fluctuation exceeding a predetermined impact fluctuation value over a duration exceeding a predetermined impact fluctuation duration;
generate at least one event-dependent products training dataset based at least in part on the product information and the historical event information;
wherein the at least one event-dependent products training dataset comprises the at least one impacted product of the plurality of products and at least one event of the plurality of historical events that is associated with at least one date of the impact;
determine an impact period for each event of the plurality of historical events based on a duration of the impact following the date of each event;
train an attribute prediction machine learning model based at least in part on the at least one event-dependent products training dataset to obtain a trained attribute prediction machine learning model;
wherein the trained attribute prediction machine learning model comprises model parameters trained to correlate each event of the plurality of historical events to:
i) the impact on each product of the plurality of products, and
ii) the impact duration of each product of the plurality of products;
collect additional event information from the one or more online sources;
wherein the additional event information is associated with one or more additional events;
wherein the additional event information comprises a date associated with each event of the one or more additional events;
apply the trained attribute prediction machine learning model to the additional event information to predict, for at least one product of the plurality of products, an additional impact and an additional impact period associated with the impact for each additional event of the at least one additional event;
wherein the additional impact comprises:
(a) a future attribute estimate,
(b) a future attribute change estimate, or
(c) both (a) and (b); and
cause to display at least one indication representative of (a) the future attribute estimate, (b) the future attribute change estimate, or (c) both (a) and (b), on a screen.
