Combined predictions methodology
A news feed system provided with an on-line social network system determines that a news feed is to be constructed for a viewer. The news feed system accesses the viewer's profile and other information associated with the viewer, accesses an inventory of activities that have been identified as potentially of interest to the viewer, and calculates relevance score for each item inventory of activities using the combined predictions methodology. The activities are then arranged for presentation to the viewer via a news feed web page, using respective calculated relevance scores.
1. A computer-implemented method comprising:
in an on-line social network system, maintaining a first statistical model for estimating probability of a click action and a second statistical model for estimating probability of a viral action;
assigning a first model weight to the first statistical model and a second model weight to the second statistical model, the first model weight and the second model weight indicate respective importance of the click action and the viral action;
for a viewer in the on-line social network system, estimating a first probability using the first statistical model, the first probability is probability that the click action will be performed by the viewer on a visual object, the visual object representing an item from a first inventory of updates identified by the on-line social network system as potentially of interest to the viewer;
estimating a second probability using the second statistical model, the second probability is probability that the viral action will be performed by the viewer on the visual object, the viral action results in an additional item being included in a second inventory of updates identified by the on-line social network system as potentially of interest to another member in the on-line social network system, the estimating of the first probability and the second probability performed prior to presenting the visual object to the viewer;
using at least one processor, calculating a relevance score for the item with respect to the viewer, by aggregating the first probability weighted by the first model weight and the second probability weighted by the first model weight; and
using the relevance score generated for the item to determine a position of the visual object in news feed relative to other visual objects in the news feed.
2. The method of claim 1 , comprising constructing a news feed web page to permit the viewer to view the news feed, the position of the visual object in the news feed determined based on the relevance score generated for the item.
3. The method of claim 1 , comprising causing presentation of the news feed web page on a display device of the viewer.
4. The method of claim 1 , wherein the click action causes presentation of further details associated with the item.
5. The method of claim 1 , wherein the viral action comprises activating a visual control presented in the news feed as associated with the visual object.
6. The method of claim 1 , wherein the calculating of the relevance score for the item comprises calculating the sum of the first probability weighted by the first model weight and the second probability weighted by the second model weight.
7. The method of claim 1 , further comprising deduplicating feature transformations prior to the generating of the estimating of the first probability and the second probability.
8. The method of claim 1 , wherein the first statistical model is logistic regression and the second statistical model is random forest.
9. The method of claim 1 , wherein the first statistical model and the second statistical model use the same statistical approach.
10. A computer-implemented system comprising:
one or more processors; and
a non-transitory computer readable storage medium comprising instructions that when executed by the one or processors cause the one or more processors to perform operations comprising:
in an on-line social network system, maintaining a first statistical model for estimating probability of a click action and a second statistical model for estimating probability of a viral action;
assigning a first model weight to the first statistical model and a second model weight to the second statistical model, the first model weight and the second model weight indicate respective importance of the click action and the viral action;
for a viewer in the on-line social network system, estimating a first probability using the first statistical model, the first probability is probability that the click action will be performed by the viewer on a visual object, the visual object representing an item from a first inventory of updates identified by the on-line social network system as potentially of interest to the viewer;
estimating a second probability using the second statistical model, the second probability is probability that the viral action will be performed by the viewer on the visual object, the viral action results in an additional item being included in a second inventory of updates identified by the on-line social network system as potentially of interest to another member in the on-line social network system, the estimating of the first probability and the second probability performed prior to presenting the visual object to the viewer;
calculating a relevance score for the item with respect to the viewer, by aggregating the first probability weighted by the first model weight and the second probability weighted by the first model weight; and
using the relevance score generated for the item to determine a position of the visual object in news feed relative to other visual objects in the news feed.
11. The system of claim 10 , comprising constructing a news feed web page to permit the viewer to view the news feed, the position of the item in the news feed determined based on the relevance score generated for the visual object.
12. The system of claim 10 , comprising causing presentation of the news feed web page on a display device of the viewer.
13. The system of claim 10 , wherein the click action causes presentation of further details associated with the item.
14. The system of claim 10 , wherein the viral action comprises activating a visual control presented in the news feed as associated with the visual object.
15. The system of claim 10 , comprising calculating the relevance score for the item by calculating the sum of the first probability weighted by the first model weight and the second probability weighted by the second model weight.
16. The system of claim 10 , further comprising deduplicating feature transformations prior to the estimating of the first probability and the second probability.
17. The system of claim 10 , wherein the first statistical model is logistic regression and the second statistical model is random forest.
18. The system of claim 10 , wherein the first statistical model and the second statistical model use the same statistical approach.
19. A machine-readable non-transitory storage medium having instruction data executable by a machine to cause the machine to perform operations comprising:
in an on-line social network system, maintaining a first statistical model for estimating probability of a click action and a second statistical model for estimating probability of a viral action;
assigning a first model weight to the first statistical model and a second model weight to the second statistical model, the first model weight and the second model weight indicate respective importance of the click action and the viral action;
for a viewer in the on-line social network system, estimating a first probability using the first statistical model, the first probability is probability that the click action will be performed by the viewer on a visual object, the visual object representing an item from a first inventory of updates identified by the on-line social network system as potentially of interest to the viewer;
estimating a second probability using the second statistical model, the second probability is probability that the viral action will be performed by the viewer on the visual object, the viral action results in an additional item being included in a second inventory of updates identified by the on-line social network system as potentially of interest to another member in the on-line social network system, the estimating of the first probability and the second probability performed prior to presenting the visual object to the viewer;
calculating a relevance score for the item with respect to the viewer, by aggregating the first probability weighted by the first model weight and the second probability weighted by the first model weight; and
using the relevance score generated for the item to determine a position of the visual object in news feed relative to other visual objects in the news feed.