Social media pricing engine
Processes for determining predicted prices of social media actions from disparate social media types are disclosed. In one example process, campaign attribute data associated with an advertising campaign on a social media service may be received. The campaign attribute data and historical campaign attribute data associated with previously performed advertising campaigns may be used to generate a plurality of predicted prices for the social media action. A weighted sum of the plurality of predicted prices may be calculated to generate a final predicted price for the social media action. Predicted prices for other social media actions performed on other social media services may similarly be generated. Systems and non-transitory computer-readable storage media for performing these processes are also provided.
1. A computer-implemented method for determining a price per action in a social media service, the method comprising:
receiving, at a pricing server, campaign attribute data associated with a first advertising campaign on a first social media service;
generating, by the pricing server, a first plurality of price predictions for a first social media action of the first social media service based on the campaign attribute data associated with the first advertising campaign and historical campaign data associated with a plurality of historical advertising campaigns, wherein generating the first plurality of price predictions for the first social media action of the first social media service comprises:
generating a first price prediction for the first social media action by identifying a first matching historical advertising campaign of the plurality of historical advertising campaigns using a k-Nearest Neighbor algorithm;
generating a second price prediction for the first social media action by identifying a second matching historical advertising campaign of the plurality of historical advertising campaigns using a Random Forest algorithm; and
generating a third price prediction for the first social media action by:
identifying a linear relationship between an attribute of the historical advertising campaigns and a price of the first social media action using a Multivariate Adaptive Regression Splines algorithm; and
determining the third price prediction based on a value of the attribute of the first advertising campaign and the identified linear relationship;
generating, by the pricing server, a first final price prediction for the first social media action based on the first plurality of price predictions; and
transmitting, by the pricing server, the first final price prediction.
2. The method of claim 1 , wherein the method further comprises receiving, by the pricing server, the historical campaign data associated with the plurality of historical advertising campaigns, wherein the historical campaign data comprises a plurality of price per actions and a plurality of sets of campaign attribute data associated with the plurality of historical advertising campaigns.
3. The method of claim 1 , further comprising:
performing, by the pricing server, a first set of mathematical transformations on the campaign attribute data associated with the first advertising campaign; and
performing, by the pricing server, a second set of mathematical transformations on the historical campaign data associated with the plurality of historical advertising campaigns.
4. The method of claim 1 , wherein:
the first price prediction comprises a price of a social media action in the first matching historical advertising campaign;
the second price prediction comprises a price of a social media action in the second matching historical advertising campaign; and
the third price prediction comprises a price calculated from the identified linear relationship and the value of the attribute of the first advertising campaign.
5. The method of claim 1 , wherein generating the first final price prediction for the first social media action based on the first plurality of price predictions comprises:
assigning a first weight to the first price prediction;
assigning a second weight to the second price prediction;
assigning a third weight to the third price prediction; and
generating the final price prediction by calculating a weighted sum of the first plurality of price predictions using the first weight, the first price prediction, the second weight, the second price prediction, the third weight, and the third price prediction.
6. The method of claim 5 , wherein the first weight, the second weight, and the third weight are generated using a root-mean square error algorithm.
7. The method of claim 1 , wherein the campaign attribute data associated with the first advertising campaign is received from a user requesting the first final predicted price, the first social media service, or a third party aggregator of advertising data.
8. The method of claim 1 , further comprising:
receiving, at the pricing server, campaign attribute data associated with a second advertising campaign on a second social media service, wherein the first social media service and the second social media service are different;
generating, by the pricing server, a second plurality of price predictions for a second social media action of the second social media service, wherein the first social media action and the second social media action are different;
generating, by the pricing server, a second final price prediction for the second social media action based on the second plurality of price predictions; and
transmitting, by the pricing server, the second final price prediction.
9. A system for determining a price per action in a social media service, the system comprising:
a pricing server configured to:
receive campaign attribute data associated with a first advertising campaign on a first social media service;
generate a first plurality of price predictions for a first social media action of the first social media service based on the campaign attribute data associated with the first advertising campaign and historical campaign data associated with a plurality of historical advertising campaigns, wherein generating the first plurality of price predictions for the first social media action of the first social media service comprises:
generating a first price prediction for the first social media action by identifying a first matching historical advertising campaign of the plurality of historical advertising campaigns using a k-Nearest Neighbor algorithm;
generating a second price prediction for the first social media action by identifying a second matching historical advertising campaign of the plurality of historical advertising campaigns using a Random Forest algorithm; and
generating a third price prediction for the first social media action by:
identifying a linear relationship between an attribute of the historical advertising campaigns and a price of the first social media action using a Multivariate Adaptive Regression Splines algorithm; and
determining the third price prediction based on a value of the attribute of the first advertising campaign and the identified linear relationship;
generate a first final price prediction for the first social media action based on the first plurality of price predictions; and
transmit the first final price prediction.
10. The system of claim 9 , wherein the pricing server is further configured to receive the historical campaign data associated with the plurality of historical advertising campaigns, wherein the historical campaign data comprises a plurality of price per actions and a plurality of sets of campaign attribute data associated with the plurality of historical advertising campaigns.
11. The system of claim 9 , wherein the pricing server is further configured to:
perform a first set of mathematical transformations on the campaign attribute data associated with the first advertising campaign; and
perform a second set of mathematical transformations on the historical campaign data associated with the plurality of historical advertising campaigns.
12. The system of claim 9 , wherein the pricing server is further configured to:
receive campaign attribute data associated with a second advertising campaign on a second social media service, wherein the first social media service and the second social media service are different;
generate a second plurality of price predictions for a second social media action of the second social media service, wherein the first social media action and the second social media action are different;
generate a second final price prediction for the second social media action based on the second plurality of price predictions; and
transmit the second final price prediction.
13. A non-transitory computer-readable storage medium having computer-executable instructions for determining a price per action in a social media service, the instructions for:
receiving, at a pricing server, campaign attribute data associated with a first advertising campaign on a first social media service;
generating, by the pricing server, a first plurality of price predictions for a first social media action of the first social media service based on the campaign attribute data associated with the first advertising campaign and historical campaign data associated with a plurality of historical advertising campaigns, wherein generating the first plurality of price predictions for the first social media action of the first social media service comprises:
generating a first price prediction for the first social media action by identifying a first matching historical advertising campaign of the plurality of historical advertising campaigns using a k-Nearest Neighbor algorithm;
generating a second price prediction for the first social media action by identifying a second matching historical advertising campaign of the plurality of historical advertising campaigns using a Random Forest algorithm; and
generating a third price prediction for the first social media action by:
identifying a linear relationship between an attribute of the historical advertising campaigns and a price of the first social media action using a Multivariate Adaptive Regression Splines algorithm; and
determining the third price prediction based on a value of the attribute of the first advertising campaign and the identified linear relationship;
generating, by the pricing server, a first final price prediction for the first social media action based on the first plurality of price predictions; and
transmitting, by the pricing server, the first final price prediction.
14. The non-transitory computer-readable storage medium of claim 13 , further comprising instructions for receiving, by the pricing server, the historical campaign data associated with the plurality of historical advertising campaigns, wherein the historical campaign data comprises a plurality of price per actions and a plurality of sets of campaign attribute data associated with the plurality of historical advertising campaigns.
15. The non-transitory computer-readable storage medium of claim 13 , further comprising instructions for:
performing, by the pricing server, a first set of mathematical transformations on the campaign attribute data associated with the first advertising campaign; and
performing, by pricing server, a second set of mathematical transformations on the historical campaign data associated with the plurality of historical advertising campaigns.
16. The non-transitory computer-readable storage medium of claim 13 , wherein generating the first final price prediction for the first social media action based on the first plurality of price predictions comprises:
assigning a first weight to the first price prediction;
assigning a second weight to the second price prediction;
assigning a third weight to the third price prediction; and
generating the final price prediction by calculating a weighted sum of the first plurality of price predictions using the first weight, the first price prediction, the second weight, the second price prediction, the third weight, and the third price prediction.
17. The non-transitory computer-readable storage medium of claim 13 , further comprising instructions for:
receiving, at the pricing server, campaign attribute data associated with a second advertising campaign on a second social media service, wherein the first social media service and the second social media service are different;
generating, by the pricing server, a second plurality of price predictions for a second social media action of the second social media service, wherein the first social media action and the second social media action are different;
generating, by the pricing server, a second final price prediction for the second social media action based on the second plurality of price predictions; and
transmitting, by the pricing server, the second final price prediction.