Machine learning based content delivery

• US 10,311,371 B1
• Filed: 12/19/2014
• Issued: 06/04/2019
• Est. Priority Date: 12/19/2014
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Hotchkies, Blair Livingstone, Bowman, Bradley Scott, Cerda, Paul Christopher, Chong, Min, Chor, Anthony T., Dirac, Leo Parker, Granade, Kevin Andrew, Pant, Udip, Scott, Sean Michael, Agarwal, Aman
• Primary Examiner(s): Cassity, Robert A
• Assistant Examiner(s): Coughlan, Peter D
• Application Number:
  US14/578,050
• Time in Patent Office: 1,628 Days
• CPC Class Codes: G06F 18/23 Clustering techniques
• CPC Class Codes: G06F 18/24 Classification techniques
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 5/01 Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
• CPC Class Codes: G06N 5/02 Knowledge representation; Symbolic representation
• CPC Class Codes: H04L 67/02 based on web technology, e.g. hypertext transfer protocol [HTTP]
• CPC Class Codes: H04L 67/1097 for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS]
• CPC Class Codes: H04L 67/568 Storing data temporarily at an intermediate stage, e.g. caching
• CPC Class Codes: H04L 67/5681 Pre-fetching or pre-delivering data based on network characteristics

Abstract

Systems and methods for managing content delivery functionalities based on machine learning models are provided. In one aspect, content requests are routed in accordance with clusters of historical content requests to optimize cache performance. In another aspect, content delivery strategies for responding to content requests are determined based on a model trained on data related to historical content requests. The model may also be used to determine above-the-fold configurations for rendering responses to content requests. In some embodiments, portions of the model can be executed on client computing devices.

Claims

1. A computer-implemented method for facilitating content delivery, the computer-implemented method comprising:

under control of a hardware computing device configured with specific computer executable instructions,

obtaining first data characterizing individual requests of a plurality of content requests over a specified period of time;

obtaining second data characterizing content delivery strategy in response to individual requests of the plurality of content requests;

obtaining third data characterizing content delivery performance in response to individual requests of the plurality of content requests;

training a machine learning model that predicts content delivery performance, wherein the machine learning model is trained, at least in part, on the first, second and third data, and wherein the machine learning model is structured such that a portion of the machine learning model can be executed by one or more user computing devices;

obtaining an incoming content request from a first user computing device;

determining a predicted content delivery performance for configuring a requested content page in a first above-the-fold (ATF) configuration in response to the incoming content request based, at least in part, on the machine learning model; and

in response to a determination that the predicted content delivery performance satisfies a predetermined condition;

configuring the requested content page in the first ATF configuration for display in a user interface;

generating a response to the first content request in accordance with configuring the requested content page in the first ATF configuration, wherein the generated response indicates a priority for retrieving network resources to be displayed in an ATF portion of the requested content page; and

transmitting the generated response to the user computing device, wherein reception of the generated response causes the user computing device to retrieve the network resources in accordance with the priority indicated in the generated response.

2. The computer-implemented method of claim 1, wherein the first data includes at least one of a type of requested resource, request timing information, associated network condition or topology, characteristics of requesting devices, or associated location information.

3. The computer-implemented method of claim 1, wherein the second data includes at least one of inclusion or exclusion of features, lazy-loading or prefetching of resources, in-lining or external calls for resources, low quality or high quality data formats, associated dependency graphs, above-the-fold information, or request routing information.

4. The computer-implemented method of claim 1, wherein the third data includes at least one of a total time to load a content page or individual network resources, number of times the content page or individual network resources was retrieved, bandwidth utilization, network latency, number of hops between client and server, processor utilization, memory utilization, cache hit or miss ratio, or load time per cache miss.

5. The computer-implemented method of claim 1, wherein the predicted content delivery performance includes a predicted valuation of a user associated with the incoming content request.

6. A computer-implemented method comprising:

under control of a hardware computing device configured with specific computer executable instructions,

obtaining a machine learning model trained to predict content delivery performance based on information related to content requests and content delivery strategies;

generating a prediction of content delivery performance by providing the obtained machine learning model with at least one attribute of a first content request or at least one portion of a first content delivery strategy for configuring a content page corresponding to the first content request in a first above-the-fold (ATF) configuration for display in a user interface; and

in response to a determination that the prediction of content delivery performance satisfies a predetermined condition, configuring the content page in the first ATF configuration and indicating a priority for retrieving network resources to be displayed in an ATF portion of the content page.

7. The computer-implemented method of claim 6, wherein the first content delivery strategy further includes at least one of inclusion or exclusion of a feature, lazy-loading or prefetching of a resource, in-lining or external call for a resource, data format, dependency graph, or routing of a request.

8. The computer-implemented method of claim 6, wherein the at least one attribute of the first content request corresponds to information related to a user associated with the first content request.

9. The computer-implemented method of claim 8, wherein the information related to the user includes at least one of user demographics, cultural preferences, geographic location, occupation, income, spending levels, content interests, hobbies, preferences, settings, purchase histories, Web browsing histories, search histories, session tracking histories, user interaction data, ownership or rental lists, or user reviews.

10. The computer-implemented method of claim 6 further comprising obtaining content delivery performance data that corresponds to the implementation of the first content delivery strategy in response to the first content request.

11. The computer-implemented method of claim 10 further comprising causing updating of the machine learning model based, at least in part, on the obtained content delivery performance data.

12. A system comprising:

a data store configured to at least store computer-executable instructions; and

a hardware processor in communication with the data store, the hardware processor configured to execute the computer-executable instructions to at least;

obtain a machine learning model trained to predict content delivery performance based on information related to content requests and content delivery strategies;

generate a first prediction of content delivery performance by providing the obtained machine learning model with at least one attribute of a first content request or at least one portion of a first content delivery strategy for configuring a content page corresponding to the first content request in a first above-the-fold (ATF) configuration for display in a user interface; and

in response to a determination that the first prediction of content delivery performance satisfies a predetermined condition, configure the content page in the first ATF configuration and indicating a priority for retrieving network resources to be displayed in an ATF portion of the content page.

13. The system of claim 12, wherein the machine learning model is trained on historical data regarding content requests over a specified period of time.

14. The system of claim 13, wherein the hardware process is further configured to at least determine the first content delivery strategy based, at least in part, on a correlation between content requests and content delivery strategies derived from the historical data.

15. The system of claim 12, wherein the hardware processor is further configured to at least generate a second prediction of content delivery performance by providing the obtained machine learning model with at least one portion of a second content delivery strategy.

16. The system of claim 15, wherein the predetermined condition includes a comparison between the first and second predictions.

17. A non-transitory computer readable storage medium storing computer executable instructions that when executed by a processor perform operations comprising:

obtaining a machine learning model trained to predict content delivery performance based on information related to content requests and content delivery strategies;

generating a prediction of content delivery performance by providing the obtained machine learning model with at least one attribute of a first content request or at least one portion of a first content delivery strategy for configuring a content page corresponding to the first content request in a first above-the-fold (ATF) configuration for display in a user interface; and

in response to a determination that the prediction of content delivery performance satisfies a predetermined condition, configuring the content page in the first ATF configuration and indicating a priority for retrieving network resources to be displayed in an ATF portion of the content page.

18. The non-transitory computer-readable storage medium of claim 17, wherein the machine learning model corresponds to a decision tree model or an artificial neural network model.

19. The non-transitory computer-readable storage medium of claim 17, wherein implementing the first content delivery strategy further comprises at least one of identifying a pre-generated response, dynamically generating HTML documents, or routing the first content request to a corresponding content serving device, in accordance with the first content delivery strategy.

20. The non-transitory computer-readable storage medium of claim 17, wherein the operations further comprise obtaining an updated machine learning model, wherein the updated machine learning model is trained on at least a portion of content delivery performance data corresponding to the implementation of the first content delivery strategy in response to the first content request.

21. The non-transitory computer-readable storage medium of claim 20, wherein the operations further comprise implementing a second content delivery strategy in response to a second content request based, at least in part, on the updated machine learning model.