METHOD AND APPARATUS FOR PREDICTING VALUE OF CACHING CONTENT ITEMS

• US 20100005046A1
• Filed: 07/02/2008
• Published: 01/07/2010
• Est. Priority Date: 07/02/2008
• Status: Active Grant

• Current Assignee: Alcatel-Lucent SA (Nokia Corporation)
• Original Assignee: Alcatel-Lucent SA (Nokia Corporation)
• Inventors: Segel, Jonathan
• Granted Patent: US 9,178,959 B2
• Time in Patent Office: Days
• US Class Current: 706/46
• CPC Class Codes: H04L 67/56 Provisioning of proxy services store-and-forward switching systems in data switching networks H04L12/54
• CPC Class Codes: H04L 67/5681 Pre-fetching or pre-delivering data based on network characteristics

Abstract

The invention includes a method and apparatus for predicting value of caching content items. In one embodiment, a method for selecting ones of a plurality of content items to be cached includes, for each of the content items, determining a probability of use of the content item comprising a probability that the content item will be used if cached, determining a value of caching the content item, and computing a caching value estimate associated with the content item using the probability of use of the content item and the value of caching the content item. The content items to be cached may then be selected using the respective caching value estimates of the content items.

Claims

1. A method for selecting ones of a plurality of content items to be cached, comprising:

determining, for each of the content items, a probability of use of the content item comprising a probability that the content item will be used if cached;

determining, for each of the content items, a value of caching the content item;

computing, for each of the content items, a caching value estimate associated with the content item using the probability of use of the content item and the value of caching the content item; and

selecting ones of the content items to be cached at least using the respective caching value estimates of the content items.

2. The method of claim 1, wherein, for each content item, the probability of use of the content item is one of a posterior probability of use estimate that is calculated using a Bayesian updating process and a directly observable probability of use estimate given a set of user signaled information.

3. The method of claim 2, wherein, when the probability of use for each content item is a posterior probability of use estimate that is calculated using a Bayesian updating process, the posterior probability of use estimate for the content item is computed using:

a prior probability of the content item being used in the absence of user signaled information;

a predicted signaled factor probability; and

a joint probability of receiving the user signaled information in combination with the content item being used by the user.

4. The method of claim 3, wherein the prior probability that the content item will be used in the absence of user signaled information is determined using at least one of a past content viewing history of the user, content preference information associated with the user, and information indicative of popularity of the content item with at least one other user.

5. The method of claim 2, wherein, when the probability of use for each content item is a directly observable probability of use estimate given a set of user signaled information, the probability of use estimate for the content item is computed using:

a predicted signaled factor probability; and

a joint probability of receiving the user signaled information in combination with the content item being used by the user.

6. The method of claim 1, wherein, for each content item, the probability of use for the content item is computed using predicted signaled factor probability, wherein the predicted signaled factor probability is determined by:

receiving previous signaled factors associated with the user, wherein the previous signaled factors comprise at least one signaled factor entered by the user via at least one user device in the past; and

determining the predicted signaled factor probability using the at least one user signaled factor, wherein the predicted signaled factor probability comprises at least one predicted signaled factor.

7. The method of claim 6, wherein the at least one predicted signaled factor comprises at least one of an audience factor reflecting which individuals will be consuming the content, a genre factor, an available-time factor reflecting available time to consume content items, a factor reflecting the preferred length of content items, a preferred content item mood factor, a popularity factor, a critical rating factor, a factor reflecting audience age, a factor for signaling potentially offensive content, and a factor reflecting preferred tempo of the content item.

8. The method of claim 6, wherein at least one of the predicted signaled factors is predicted as a function of at least one external factor.

9. The method of claim 8, wherein the at least one external factor comprises at least one time factor of time of day, day of week, day of year, and specified holiday.

10. The method of claim 1, wherein, for each content item, the probability of use of the content item is calculated as a sum, across a plurality of combinations of user signaled information, of a probability of use of the content item within that combination multiplied by a probability of occurrence of that combination.

11. The method of claim 1, wherein, for each content item, the probability of use of the content item is calculated as a time-weighted sum of probability of use of the content item at one or more points in the future.

12. The method of claim 1, wherein, for each content item, determining the value of caching the content item comprises:

determining a benefit of caching the content item;

determining a cost of caching the content item; and

determining the value of caching the content item using the benefit of caching the content item and the cost of caching the content item.

13. The method of claim 12, where the benefit of caching is:

lowered if the content item is already cached; and

increased if caching is likely to de-load the network during a peak usage period.

14. The method of claim 12, where the cost of caching is dependent on at least one of a size of the content item, available space in a cache, and network resources that are consumed to retrieve the content item.

15. The method of claim 1, wherein, for each of the content items, the value of caching the content item comprises a weighting factor.

16. The method of claim 15, wherein the weighting factor is dependent on at least one of a network loading factor, a network overloading cost factor, and a size of the content item.

17. The method of claim 1, wherein, for each of the content items, the caching value estimate associated with the content item is computed as:

${\mathrm{value}}_k={\int <br><br>}_{\mathrm{now}}^{\mathrm{\infty <br><br>}}<br><br>\mathrm{prob}<br><br><br><br>\left(I_{k,t}\right)<br><br>W_{t,k}<br><br><br><br><br><br>t,$

wherein prob(I_k,t) comprises the probability of use of the content item k at a time t, wherein W_t,k comprises the value of caching the content item k at time t.

18. The method of claim 17, wherein prob(I_k,t) comprises one of a cumulative posterior probability computed using a Bayesian updating approach and a cumulative probability computed using a directly-observable approach;

wherein, when the prob(I_k,t) comprises the cumulative posterior probability computed using a Bayesian updating approach, the prob(I_k,t) is computed as;

$\mathrm{prob}<br><br><br><br>\left(I_{k,t}\right)=\underset{n=1}{\overset{m}{\sum <br><br>}}<br><br>\left[\mathrm{prob}<br><br><br><br>\left(S_n|I_k\right)*\mathrm{prior}<br><br><br><br>\left(I_k\right)\right]/\mathrm{prob}<br><br><br><br>\left(S_n\right);<br><br>$

wherein, when the prob(I_k,t) comprises the cumulative probability computed using a directly-observable approach, the prob(I_k,t) is computed as

$\mathrm{prob}<br><br><br><br>\left(I_{k,t}\right)=\underset{n=1}{\overset{m}{\sum <br><br>}}<br><br>\left[\mathrm{prob}<br><br><br><br>\left(I_k|S_n\right)*\mathrm{prob}<br><br><br><br>\left(S_n\right)\right].$

19. The method of claim 1, further comprising:

propagating the selected ones of the content items toward a content cache adapted to cache the selected ones of the content items.

20. The method of claim 19, wherein the cache is located at least one of:

on a user device being used by the user; and

on a network device serving the user and at least one additional user.

21. The method of claim 1, further comprising:

propagating the caching value estimates toward at least one element, wherein the at least one element controls selecting the ones of the content items to be cached at least based on the caching value estimates.

22. An apparatus for selecting ones of a plurality of content items to be cached, comprising:

means for determining, for each of the content items, a probability of use of the content item comprising a probability that the content item will be used if cached;

means for determining, for each of the content items, a value of caching the content item;

means for computing, for each of the content items, a caching value estimate associated with the content item using the probability of use of the content item and the value of caching the content item; and

means for selecting ones of the content items to be cached at least using the respective caching value estimates of the content items.

23. A computer readable medium storing a software program which, when executed by a computer, causes the computer to perform a method for selecting ones of a plurality of content items to be cached, the method comprising:

determining, for each of the content items, a probability of use of the content item comprising a probability that the content item will be used if cached;

determining, for each of the content items, a value of caching the content item;

computing, for each of the content items, a caching value estimate associated with the content item using the probability of use of the content item and the value of caching the content item; and

selecting ones of the content items to be cached at least using the respective caching value estimates of the content items.