US 10,949,752 B1Granted

System and method of portfolio matching

AssigneeApplied Predictive Technologies, Inc.

InventorsBradley Philip Burns, Anthony Dean Bruce, Arvind Bhusnurmath

Life of this patent

Estimating the term…

Loading patent…

Abstract

The embodiments the systems and methods described herein attempt to optimally select a group or portfolio of control locations for each test location. The optimization can be generally performed in two steps. First, an objective function is defined that scores the similarity of a set of control locations averaged together. Second, given the large number of potential solutions, a computationally-feasible algorithm that identifies an optimal set of control locations and is based on the objective function is executed. In order to obtain the optimal set of control locations in an efficient manner for use in business analytics, the algorithm may use a hill-climbing algorithm. As a result, an optimization function can be incrementally improved in an efficient manner.

Claims16

Dependent claims

A computer-implemented method of reducing processing power used by a computer when identifying at least 10 control locations out of a set of at least 500 control locations for a particular test location, the method comprising:

generating, in a database by the computer, a cohort data set of a subset of control locations from the set of control locations;
determining, by the computer, a first deviation of a performance trend of the cohort data set of the control locations from a performance trend of the test location;
randomly parsing, by the computer, one or more control locations from the cohort data set;
randomly parsing, by the computer, one or more control locations from the set of control locations not in the cohort data set;
generating, by the computer, a revised cohort data set by replacing the selected one or more control locations from the cohort data set;
determining, by the computer, a second deviation of a performance trend of the revised cohort data set from the performance trend of the test location;
determining, by the computer, whether the second deviation is smaller than the first deviation;
when the second deviation is larger than the first deviation, incrementing, by the computer, a counter;
reiteratively generating, by the computer, a new revised cohort data set by replacing one or more control locations of the revised cohort data set until the counter reaches a predetermined threshold value;
and
upon the counter reaching the predetermined threshold value, generating, by the computer, a final cohort data set with the subset of control locations having the smallest deviation of the performance trend and triggering a server to perform a test versus control analysis using the generated final cohort.

8 more elements

A computer-implemented method of reducing processing power used by a computer when identifying at least 10 control locations out of a set of at least 500 control locations most similar to particular test location data, the method comprising:

generating, in a database by the computer, a baseline cohort data set of a subset of control locations from the set of control locations;
calculating, by the computer, a similarity between the test location data and the baseline cohort data set;
randomly parsing, by the computer, one or more control locations from the baseline cohort data set;
randomly parsing, by the computer, one or more control locations from the set of control locations not in the baseline cohort data set and replacing the selected one or more control locations from the baseline cohort data set to form a revised cohort data set;
calculating, by the computer, a similarity between the test location data and the revised cohort data set;
determining, by the computer, whether the revised cohort data set is more similar to the test location data than the baseline cohort data set based upon calculating the similarity with the test location data;
reiteratively generating, by the computer, a new revised cohort data set by replacing one or more control locations of the revised cohort data set until a counter reaches a predetermined threshold value;
and
upon the counter reaching the predetermined threshold value, generating, by the computer, a final cohort data set with the subset of control locations having the smallest deviation of the performance trend and triggering a server to perform a test versus control analysis using the generated final cohort.

6 more elements

A computer system for reducing processing power when selecting a control portfolio, the system comprising:

a computer configured to:
generate in a database a data set representing a subset of control locations from the set of at least 500 control locations;
determine a first similarity measure of historical performance of one or more criteria between the data set representing the subset of control locations and the test location over a period of time;
randomly parse one or more control locations from the data set representing the subset of control locations;
select one or more control locations from the set of control locations not in the data set representing the subset of control locations;
replace the one or more selected control locations from the subset with the one or more control locations not in the subset to form a revised data set representing subset of control locations;
determine a second similarity measure of historical performance of one or more criteria between the revised data set representing subset and the test location over the period of time;
determine whether the revised data set representing subset is more similar to the test location than the subset based upon the first and second similarity measures;
when the second similarity measure is larger than the first similarity measure, incrementing a counter;
reiteratively generating a new data set by replacing one or more control locations of the revised data set until the counter reaches a predetermined threshold value;
and
upon the counter reaching the predetermined threshold value, generating a final data set with the subset of control locations having the smallest deviation of the similarity measure and triggering a server to perform a test versus control analysis using the generated final data set, wherein the final dataset comprises at least 1
0 control locations out of the set of at least 500 control locations.

10 more elements

Dependencies are inferred from the claim text.