Using media events to predict time series data

• US 9,824,146 B1
• Filed: 05/17/2012
• Issued: 11/21/2017
• Est. Priority Date: 05/17/2012
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Siddiqui, Muhammad Ali, Bodell, Colin, Bilger, Jeff M.
• Primary Examiner(s): Hill, Stanley K
• Assistant Examiner(s): Coughlan, Peter
• Application Number:
  US13/474,037
• Time in Patent Office: 2,014 Days
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities
• CPC Class Codes: G06Q 30/0601 Electronic shopping [e-shopping]

Abstract

Disclosed are various embodiments for using media reported events to generate predictions for time series data. Information retrieved from a plurality of network content sources is classified into a plurality of categories. A prediction is generated for a time series. The time series is associated with a metric observed in a computing system. The generated prediction takes into account an impact of at least one of instance of the classified information.

Claims

1. A non-transitory computer-readable medium embodying a program executable in a computing device, wherein when executed, the program causes the computing device to at least:

crawl a plurality of network sites to retrieve a plurality of network pages;

parse the plurality of network pages to discover a plurality of media reported events;

index the plurality of media reported events;

classify the indexed media reported events into a plurality of categories;

receive a request for a prediction for a time series;

generate, in response to the request, an initial prediction for the time series, the initial prediction being independent of the plurality of media reported events, the time series associated with a metric observed in a computing system;

measure an effect of at least one of the plurality of media reported events on the time series;

revise the initial prediction to take into account the effect of the at least one of the plurality of media reported events; and

respond to the request by providing the revised prediction.

2. The non-transitory computer readable medium of claim 1, wherein the plurality of network pages comprise at least one of a social networking site, a news publisher site, or a news aggregator site.

3. A method, comprising:

classifying, by a computing device, information retrieved from a plurality of network content sources into a plurality of categories;

determining, by the computing device, an impact of at least one instance of the classified information on a metric associated with a time series by determining a correlation between the at least one instance of the classified information and the metric; and

generating, by the computing device, a prediction for the time series taking into account the impact of the at least one instance of the classified information.

4. The method of claim 3, wherein the prediction is generated by at least one of statistical analysis, regression, Gaussian process regression, or machine learning.

5. The method of claim 3, further comprising:

determining, by the computing device, an accuracy of the prediction associated with the metric; and

improving, by the computing device, the accuracy of a subsequent prediction associated with the metric.

6. The method of claim 3, further comprising:

wherein the prediction is generated in response to receiving a request for the prediction.

7. The method of claim 3, wherein the correlation is determined using a machine learning technique.

8. The method of claim 3, wherein the generating further comprises:

generating, by the computing device, a first prediction for the time series that is independent of the determined impact; and

generating, by the computing device, as the prediction, a revised prediction for the time series that takes into account the impact of the at least one instance of the classified information on the metric.

9. The method of claim 3, wherein the categories are hierarchical.

10. The method of claim 3, wherein the at least one instance of the classified information is discovered through at least one of a social networking network site, a news publication network site, or an electronic commerce network site.

11. The method of claim 3, further comprising:

crawling, by the computing device, a plurality of network sites to discover the information; and

indexing, by the computing device, the information.

12. The method of claim 3, further comprising:

receiving, by the computing device, a feed from at least one network site, the feed describing the information; and

indexing, by the computing device, the information.

13. The method of claim 3, further comprising generating, by the computing device, at least one of the categories based at least in part on the information.

14. A system, comprising:

at least one computing device; and

at least one application implemented in the at least one computing device, wherein when executed, the at least one application causes the at least one computing device to at least;

classify a plurality of items of information reported from at least one network content source into a plurality of categories;

measure an effect of at least one of the reported items of information on at least one metric observed in a computing system by determining a correlation between the at least one of the reported items of information and the at least one metric;

generate an initial prediction for a time series that is associated with the at least one metric, the initial prediction being independent of the effect of the at least one of the reported items of information on the at least one metric; and

revise the initial prediction to take into account the effect of the at least one of the reported items of information on the at least one metric.

15. The system of claim 14, wherein when executed, the at least one application further causes the at least one computing device to at least apply a plurality of business rules to select a portion of the reported items of information.

16. The system of claim 14, wherein the correlation is determined using a machine learning technique.

17. The system of claim 14, wherein when executed, the at least one applicaton further causes the at least one computing device to at least assign an importance to each category.

18. The system of claim 14, wherein when executed, the at least one application further causes the at least one computing device to at least learn the categories by analyzing the reported items of information.

19. The system of claim 18, wherein learning the categories further comprises using a clustering algorithm.

20. The system of claim 14, wherein classifying a plurality of items further comprises using a machine learning technique.

21. The system of claim 20, wherein the machine learning technique is at least one of a classifier, a forest classifier, a support vector machine, a clustering analyzer, or a neural network.

22. The system of claim 14, wherein the at least one of the reported items of information is indicated in a prediction request.

23. The system of claim 14, wherein the initial prediction for the time series is generated according to a prediction model indicated in a prediction request.