Privacy compliant multiple dataset correlation system
View Patent ↗A system and method for using inverse mathematical principles in the analysis of compatible datasets so that correlations and trends within and between said datasets can be uncovered. The present invention is tailored to the analysis of datasets that are extremely large; result from passive, privacy-secure, or anonymous, data collections; and are relatively unbiased. Correlations and trends uncovered by such analysis can be further examined by data mining and prediction portions of the present invention, which uncover and make use of interrelated rules that determine data structures.
1. A method comprising:
collecting at least one dynamic dataset representing human behavior, and at least one static dataset representing human characteristics, wherein the dynamic dataset and the static dataset share at least one common characteristic and have an assumed relationship, wherein the dynamic dataset and the static dataset are collected without knowing individual-specific demographic information;
selecting with a computer processor at least one subset of the dynamic dataset and the static dataset that share at least one common characteristic;
expressing the assumed relationship between the dynamic dataset and the static dataset as a mathematical assumption;
defining an error function which describes the dynamic dataset and the static dataset in terms of the mathematical assumption;
performing at least one fitting procedure to calculate values that define the mathematical assumption;
performing at least one fitting procedure to account for errors in the assumed relationship; and
using the computer processor to store the mathematical assumption in a database as a rule system between the dynamic dataset and the static dataset.
2. The method of claim 1 , wherein the dynamic dataset corresponds to set-top box event data.
3. The method of claim 2 , wherein the static dataset corresponds to demographic data.
4. The method of claim 3 , further comprising:
correlating between the set-top box event data and the demographic data to determine the relationship of demographic data to content viewership.
5. A method comprising:
collecting at least one dynamic dataset representing human behavior, and at least one static dataset, wherein the dynamic dataset and the static dataset share at least one common characteristic and have an assumed relationship, wherein the dynamic dataset and the static dataset are collected without knowing individual-specific demographic information;
selecting with a computer processor at least one subset of the dynamic dataset and the static dataset the share at least one common characteristic;
expressing the assumed relationship as a mathematical assumption;
defining an error function which describes the dynamic dataset and the static dataset in terms of the mathematical assumption;
performing at least one fitting procedure to calculate values that define the mathematical assumption;
performing at least one fitting procedure to account for errors in the assumed relationship;
using the processor to store the mathematical assumption and the error assumption in an individual-specific array in a database; and
repeating this process, such that a plurality of mathematical assumptions and error functions are stored in individual-specific array.
6. The method of claim 5 , wherein the dynamic dataset corresponds to set-top box event data.
7. The method of claim 6 , wherein the static dataset corresponds to demographic data.
8. The method of claim 7 , wherein the individual-specific array corresponds to a privacy-compliant identification number.
9. The method of claim 8 , wherein the privacy-complaint identification number is a set-top box identification number.