Machine learning based webinterface generation and testing system

• US 10,438,111 B2
• Filed: 01/05/2017
• Issued: 10/08/2019
• Est. Priority Date: 01/05/2016
• Status: Active Grant

• Current Assignee: Evolv Technology Solutions, Inc.
• Original Assignee: Evolv Technology Solutions, Inc.
• Inventors: Iscoe, Neil, Miikkulainen, Risto
• Primary Examiner(s): Starks, Wilbert L
• Application Number:
  US15/399,433
• Publication Number:
  US 20170193403A1
• Time in Patent Office: 1,006 Days
• Field of Search: 706 13, 706 45
• CPC Class Codes: G06F 11/3664 Environments for testing or debugging software
• CPC Class Codes: G06F 16/2379 Updates performed during online database operations; commit processing
• CPC Class Codes: G06F 16/26 Visual data mining; Browsing structured data
• CPC Class Codes: G06F 16/958 Organisation or management of web site content, e.g. publishing, maintaining pages or automatic linking
• CPC Class Codes: G06F 3/0484 for the control of specific functions or operations, e.g. selecting or manipulating an object, an image or a displayed text element, setting a parameter value or selecting a range
• CPC Class Codes: G06F 40/143 Markup, e.g. Standard Generalized Markup Language [SGML] or Document Type Definition [DTD]
• CPC Class Codes: G06F 8/36 Software reuse
• CPC Class Codes: G06F 9/451 Execution arrangements for user interfaces
• CPC Class Codes: G06N 3/04 Architecture, e.g. interconnection topology
• CPC Class Codes: G06N 3/06 Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
• CPC Class Codes: G06N 3/086 using evolutionary algorithms, e.g. genetic algorithms or genetic programming
• CPC Class Codes: G06N 3/126 Evolutionary algorithms, e.g. genetic algorithms or genetic programming
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising

Abstract

Roughly described, the technology disclosed provides a so-called machine learned conversion optimization (MLCO) system that uses evolutionary computations to efficiently identify most successful webpage designs in a search space without testing all possible webpage designs in the search space. The search space is defined based on webpage designs provided by marketers. Website funnels with a single webpage or multiple webpages are represented as genomes. Genomes identify different dimensions and dimension values of the funnels. The genomes are subjected to evolutionary operations like initialization, testing, competition, and procreation to identify parent genomes that perform well and offspring genomes that are likely to perform well. Each webpage is tested only to the extent that it is possible to decide whether it is promising, i.e., whether it should serve as a parent for the next generation, or should be discarded.

Claims

1. A computer-implemented webinterface generation and testing system promoting predetermined target user behavior, comprising:

a memory storing a candidate database having a population of candidate individuals, each of the candidate individuals identifying a funnel of one or more webinterfaces, each funnel having a plurality of dimensions, the funnel having a mechanism by which achievement of the target user behavior can be detected, each candidate individual identifying a value for each of the dimensions;

a population initialization module which writes a preliminary pool of candidate individuals into the candidate individual population;

a candidate testing module that tests each of the candidate individuals in the candidate individual population by live online evaluation, the candidate testing module developing a performance measure for each of the candidate individuals undergoing the live online evaluation, the performance measure being an indication of success of the funnel in achieving the target user behavior;

a procreation module which adds to the candidate individual population new individuals formed in dependence upon a respective set of one or more parent individuals from the candidate individual population and parents'"'"' performance measures; and

a control module that iterates the candidate testing module and the procreation module.

2. The system of claim 1, wherein the procreation module forms one of the new individuals such that genes for the new individual are fitness-proportionally selected from the parent individuals in the candidate individual population.

3. The system of claim 1, further configured to include a competition module which selects individuals undergoing the live online evaluation for discarding from the candidate individual population in dependence upon the developed performance measures.

4. The system of claim 3, wherein the procreation module adds to the candidate individual population new individuals formed in dependence upon a respective set of one or more parent individuals from the candidate individual population and not yet selected for discarding.

5. The system of claim 4, wherein the procreation module forms one of the new individuals by crossing over two or more parent individuals on a gene-by-gene basis.

6. The system of claim 4, wherein the procreation module forms one of the new individuals by mutating at least one parent individual on a gene-by-gene basis.

7. The system of claim 4, wherein the procreation module forms one of the new individuals by mutation only after forming some of the new individuals by crossover.

8. The system of claim 3, wherein the control module iterates the candidate testing module, the competition module, and the procreation module.

9. The system of claim 3, wherein the selection of individuals for discarding by the competition module includes, for a subject individual, a first instance of comparing a relative performance measure of the subject individual to that of another individual in the candidate individual population.

10. The system of claim 9, wherein the relative performance measure of the subject individual is determined based on a difference between an absolute performance measure of a control individual determined in a first time period and an absolute performance measure of the subject individual determined in the first time period.

11. The system of claim 1, further configured to include an allocation module that segments the candidate individual population into multiple population groups based on a plurality of heuristics.

12. The system of claim 11, wherein the candidate testing module variedly distributes live user traffic received by a particular organization among the population groups.

13. The system of claim 1, wherein the population initialization module creates the preliminary pool of candidate individuals by randomly selecting from a candidate search space an available value for each of the dimensions and encoding the dimensions and the dimension values as binary formulations.

14. The system of claim 13, wherein the population initialization module determines that a majority of the available dimension values in the candidate search space have been initialized and, in response, randomly writes, into the candidate individual population, candidate individuals based on the available dimension values in the candidate search space not yet initialized.

15. The system of claim 1, wherein, in writing the preliminary pool of candidate individuals into the candidate individual population, the population initialization module:

obtains at least one starter funnel, the starter funnel having a plurality of starter dimensions and a set of starter dimension values for each of the starter dimensions; and

creates the preliminary pool of candidate individuals in a manner that ensures that each of the starter dimension values occurs in only one of the candidate individuals.

16. The system of claim 1, wherein a number of iterations completed by the control module is a function of one or more testing parameters specific to a particular organization seeking promotion of the target user behavior.

17. The system of claim 16, wherein a testing parameter is an amount of live user traffic received by the particular organization over a time period.

18. The system of claim 17, wherein the control module is configured to complete the number of iterations within the time period.

19. The system of claim 17, wherein the population initialization module makes a maximum size of the preliminary pool of candidate individuals a function of the amount of live user traffic.

20. The system of claim 19, wherein the population initialization module resolves a conflict between a number of starter dimensions and starter dimension values defined for the starter funnel and the maximum size of the preliminary pool of candidate individuals by:

selecting a first subset of the starter dimensions and the starter dimension values for constructing the preliminary pool of candidate individuals in a manner that ensures that the number of iterations is completed within the time period.

21. The system of claim 20, wherein the selection of the first subset is biased towards earlier specified starter dimensions and starter dimension values.

22. The system of claim 20, wherein the procreation module forms one of the new individuals in dependence upon a second subset of the starter dimensions and the starter dimension values different from the first subset.

23. The system of claim 1, wherein the target user behavior is conversion by end users.

24. The system of claim 1, wherein the performance measure developed by the candidate testing module is based on conversion rates of the candidate individuals undergoing the live online evaluation.

25. The system of claim 1, wherein the performance measure developed by the candidate testing module is based on revenue received from the candidate individuals undergoing the live online evaluation.

26. A computer-implemented webinterface generation and testing method for promoting predetermined target user behavior, comprising:

providing a memory storing a candidate database having a population of candidate individuals, each of the candidate individuals identifying a funnel of one or more webinterfaces, each funnel having a plurality of dimensions, the funnel having a mechanism by which achievement of the target user behavior can be detected, each candidate individual identifying a value for each of the dimensions;

writing a preliminary pool of candidate individuals into the candidate individual population;

testing each of the candidate individuals in the candidate individual population by live online evaluation, the candidate testing module developing a performance measure for each of the candidate individuals undergoing the live online evaluation, the performance measure being an indication of success of the funnel in achieving the target user behavior;

adding to the candidate individual population new individuals formed in dependence upon a respective set of one or more parent individuals from the candidate individual population and parents'"'"' performance measures; and

repeating the steps of testing and adding.