Personalized car recommendations based on customer web traffic

Patent metadata

Field	Value
Publication number	US-10796355-B1
Application number	US-201916728551-A
Country	US
Kind code	B1
Filing date	Dec 27, 2019
Priority date	Dec 27, 2019
Publication date	Oct 6, 2020
Grant date	Oct 6, 2020

Abstract

One or more embodiments are generally directed to techniques to provide specific vehicle recommendations. Various techniques, methods, systems, and apparatuses include utilizing user web-traffic and/or one or more tags generated by application of a machine learning model to a data source, where the data source may include language with respect to one or more automobiles or vehicles, to provide a recommendation for a particular automobile.

First claim

What is claimed is:

1. A method, comprising:
aggregating, by a processor of a host device, automobile data comprising one or more web pages accessed by a user device and one or more search queries submitted by the user device to a search engine, the aggregated automobile data associated with a plurality of automobile makes and models;
comparing, by the processor, one or more automobile terms associated with the one or more web pages and the one or more search queries of the aggregated automobile data to a plurality of automobile tags stored on the host device to identify a subset of the plurality of automobile makes and models associated with the aggregated automobile data, wherein the plurality of automobile tags are generated by the processor applying a term frequency model to a plurality of terms in a corpus of reviews of the plurality of automobiles; and
transmitting, by the processor, a suggestion utilizing the aggregated automobile data and the plurality of tags to the user device, the suggestion including at least one of i) information identifying at least one particular automobile of the subset of automobile makes and models or ii) information identifying a subset of the plurality of tags associated with the subset of automobile makes and models.

2. The method of claim 1 , wherein the suggestion is at least one of a plurality of automobile makes and models, each one of the plurality of automobile makes and models associated with a probability that the corresponding one of the plurality of automobile makes and models is a preferred choice of the user based on the aggregated automobile data.

3. The method of claim 1 , wherein the one or more web pages comprise web pages directed to automobile content, wherein the one or more search queries are related to automobiles, the method further comprising:
recording, by a web plugin of a web browser of the user device, the one or more accessed web pages and the one or more search queries as the aggregated automobile data;
providing, by the plugin, the aggregated automobile data to the host device;
receiving, by the web plugin, the suggestion from the host device;
generating, by the web plugin, a popup window comprising the suggestion; and
outputting, by the web plugin, the popup window for display.

4. The method of claim 1 , the method further comprising:
applying, by the processor, a term frequency-inverse document frequency (TF-IDF) model to a plurality of terms in the corpus of reviews, wherein applying the TF-IDF model performs a threshold computation that assigns a higher value to one or more terms of the plurality of terms with i) a lower overall-frequency within the corpus in relation to all of the plurality of automobile makes and models and ii) a higher frequency within the corpus in relation to a particular automobile make and model of the plurality of automobile makes and models.

5. The method of claim 4 , wherein only one or more terms exceeding a threshold score associated with the threshold computation form a basis for the plurality of tags.

6. The method of claim 5 , the method further comprising:
generating, by the processor a weighted score for each of the plurality of tags based on both i) a score associated with the threshold computation and ii) the comparing of the automobile terms associated with the aggregated automobile data to the plurality of automobile tags stored on the host device.

7. The method of claim 5 , wherein the transmitted suggestion is with respect to the user selecting a web page associated with particular automobile, and wherein the comparing the frequency of the one or more automobile terms associated with the aggregated automobile data to the plurality of automobile tags stored on the host device comprises:
determining, by the processor, a frequency of the one or more automobile terms discussed in relation to one or more automobile make and models of the one or more web pages; and
generating, by the processor, a weighted score for the selection of the particular automobile based on i) the determined frequency of the one or more automobile terms discussed in relation to the one or more automobile make and models of the one or more web pages and ii) the threshold computation.

8. The method of claim 6 , wherein the transmitted suggestion is with respect to the user selecting a particular automobile.

9. The method of claim 8 , wherein the transmitting the suggestion is based on the weighted score.

10. A host system, comprising:
a memory to store instructions; and
processing circuitry, coupled with the memory, operable to execute the instructions, that when executed, cause the processing circuitry to:
aggregate automobile data accessed by a user device, the aggregated automobile data to comprise one or more web pages and one or more submitted search queries;
store the aggregated automobile data;
generate a plurality of tags related to a plurality of automobile makes and models based on another data source by applying a term frequency model for comparing a frequency of a second plurality of terms in relation to each of the plurality of automobile make and models, wherein the second plurality of terms is associated with the another data source;
generate a first score for each of the plurality of make and models based on the term frequency model;
update the first scores based on the term frequency model applied to the first scores; and
transmit a suggestion to the user device for a particular automobile based on the generated plurality of tags, the updated first scores, and the data source accessed by the user.

11. The host system of claim 10 , wherein the one or more web pages are directed to automobile content including a first plurality of terms in relation to a plurality of automobile makes and models.

12. The host system of claim 10 , wherein the term frequency model is a term frequency-inverse document frequency (TF-IDF) model.

13. The host system of claim 12 , wherein the another data source includes a corpus of user automobile reviews.

14. The host system of claim 12 , wherein the another data source includes a corpus of expert automobile reviews.

15. A non-transitory computer-readable storage medium storing computer-readable program code executable by a processor to:
aggregate automobile data comprising one or more web pages accessed by a user device and one or more search queries submitted by the user device to a search engine;
compare one or more automobile terms associated with the aggregated automobile data to a plurality of automobile tags stored on a host device to identify a subset of a plurality of automobile makes and models associated with the aggregated automobile data, wherein the plurality of tags are generated from a frequency-based machine learning model applied to a corpus of automobile reviews; and
transmit an automobile suggestion to the user device utilizing the aggregated automobile data and the plurality of tags.

Assignees

• Capital One Services Llc

Inventors

• Price Micah
• Tang Qiaochu
• Hoover Jason
• Dagley Geoffrey
• Ghamsari Avid

Classifications

• G06F16/907

  Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually · CPC title

• G06F16/90324

  using system suggestions · CPC title

• G06N20/00

  Machine learning · CPC title

• G06Q30/0641

  utilising user interfaces specially adapted for shopping · CPC title

• G06Q30/0631 Primary

  Recommending goods or services · CPC title