Machine learning based system and method of calculating a match score and mapping the match score to a level

• US 10,152,695 B1
• Filed: 10/21/2014
• Issued: 12/11/2018
• Est. Priority Date: 03/15/2013
• Status: Active Grant

• Current Assignee: Elance Incorporation
• Original Assignee: Elance Incorporation
• Inventors: Chiu, Kwan-Min, Rosati, Fabio, Poon, Yuet Ping, Kamil, Megan Emily
• Primary Examiner(s): Vincent, David R
• Application Number:
  US14/520,226
• Time in Patent Office: 1,512 Days
• Field of Search: 706 15, 706 45, 706 12
• CPC Class Codes: G06F 16/24578 using ranking
• CPC Class Codes: G06F 16/248 Presentation of query results
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06Q 10/1053 Employment or hiring

Abstract

Embodiments of the present invention are directed to a system and method of calculating a match score and mapping the match score to a level. Interested contractors in a list are ranked based on a rank score that is calculated for each of the interested contractor using a set of factors. The rank score is a combination of factor scores associated with the factors. A mapping engine, implementing a machine learning model, maps each of the interested contractors to one of at least two levels based on the set of factor scores, by comparing the set of factor scores to historical data collected to determine the likelihood of the contractor having that set being chosen by the client. Any interested contractor who has been mapped to the highest level is distinguished from others in the list. The mapping engine continuously learns from the historical data to improve future mappings.

Claims

1. A system comprising:

at least one data store configured to maintain profiles of all members, a unstructured data of each of the members, historical data, and a rank model; and

at least one server computer that together is configured to;

receive from a client device a job posting submitted by a client;

receive from a plurality of contractor devices a plurality of proposals submitted by a plurality of contractors regarding the job posting;

in response to receiving the plurality of proposals,

access the at least one data store to retrieve the unstructured data that is specific to the client and to retrieve the profiles associated with the plurality of contractors who submitted the plurality of proposals;

dynamically determine, based on the unstructured data that is specific to the client, a set of parameters that is only pertinent to the client and is to be used by the rank model;

for each of the plurality of contractors, determine, based on the rank model using the set of parameters that is only pertinent to the client and based on the profile for a corresponding contractor, a rank score comprising a set of factor scores; and

present an ordered listing of the contractors according to the rank scores; and

implement a mapping engine configured to;

access the at least one data store to retrieve the historical data; and

for each of the plurality of contractors, compare the set of factor scores associated with the corresponding contractor to past sets of factor scores that are part of the historical data to determine, based on associated data that are part of the historical data and that are regarding whether respective service providers for the past sets of factor scores had been awarded jobs, a percentage of likelihood of being selected for having the set of factor scores and to, thereby, map the corresponding contractor to one of at least two levels based on the determined percentage of likelihood of being selected.

2. The system of claim 1, wherein the at least one data store maintains a machine learning model, wherein the mapping engine implements the machine learning model to continuously learn from the historical data collected by the system to improve future mappings.

3. The system of claim 1, wherein each factor score is based on, at least, a factor weight and a personal factor value, wherein the personal factor value is determined from a criteria calculation that is the same for all members, wherein the personal factor value of a contractor calculated at a first time (T1) is different from the personal factor value of the contractor calculated at a second time (T2).

4. The system of claim 1, wherein the at least one server computer that together is further configured to receive from the client device a selection, wherein information regarding the selection becomes part of the historical data.

5. The system of claim 1, wherein the unstructured data that is specific to the client includes behavioral data.

6. The system of claim 1, wherein the unstructured data includes positive user action.

7. The system of claim 1, wherein the unstructured data includes negative user action.

8. The system of claim 1, wherein the ordered listing of the contractors includes the mapping of each contractor.

9. The system of claim 1, wherein the at least one data store maintains a machine learning model configured to continuously learn from the unstructured data that is specific to the first member to determine data points that are relevant to the first member.

10. The system of claim 9, wherein each in the set of factor scores is associated with one of the data points.

11. A system comprising:

at least one data store configured to maintain profiles of all members, an evolving collection of unstructured data of each of the members, an evolving collection of historical data, and a rank model; and

at least one server computer that together is configured to;

receive from a first end device a request submitted by a first member of all members, wherein the first member is associated with a first type of member;

receive from a plurality of second end devices a plurality of responses submitted by a subset of the members regarding the request, wherein each within the subset of members is associated with a second type of member that is different from the first type of member;

in response to receiving the plurality of responses,

access the at least one data store to retrieve the unstructured data that is specific to the first member and to retrieve the profiles associated with the subset of members who submitted the plurality of responses;

dynamically determine, based on the unstructured data that is specific to the first member, a set of parameters that is only pertinent to the first member and is to be used by the rank model;

for each in subset of members, determine, based on the rank model using the set of parameters that is only pertinent to the first member and based on the profile for a corresponding member, a rank score comprising a set of factor scores; and

present an ordered listing of the members in the subset of members according to the rank scores;

implement a mapping engine configured to;

access the at least one data store to retrieve the historical data; and

for each in the subset of members, compare the set of factor scores associated with the corresponding member to past sets of factor scores that are part of the historical data to determine, based on associated data that are part of the historical data and that are regarding whether respective service providers for the past set of factor scores had been awarded jobs, a percentage of likelihood of being selected for having the set of factor scores and to, thereby, map the corresponding member to one of at least two levels based on the determined percentage of likelihood of being selected, wherein each of the at least two levels are distinguished in the ordered listing; and

receive from the first end device a selection, wherein information regarding the selection becomes part of the evolving collection of historical data.

12. The system of claim 11, wherein the at least one data store maintains a machine learning model configured to continuously learn from the unstructured data that is specific to the first member to determine data points that are relevant to the first member.

13. The system of claim 12, wherein each in the set of factor scores is associated with one of the data points.

14. The system of claim 13, wherein each factor score is based on, at least, a factor weight and a personal factor value, wherein the personal factor value is determined from a criteria calculation that is the same for all members, wherein the personal factor value of a member calculated at a first time (T1) is different from the personal factor value of that member calculated at a second time (T2).

15. The system of claim 13, wherein the machine learning model is also configured to continuously learn from the historical data collected by the system to improve future mappings.

16. The system of claim 15, wherein the unstructured data that is specific to the first member includes behavioral data.

17. The system of claim 16, wherein the ordered listing of the members in the subset of members includes the mapping of each member in the subset of members.

18. The system of claim 11, wherein the ordered listing of the members in the subset of members is displayed in a user interface, wherein each in the ordered listing includes at least one contact feature for directly contacting an associated member without leaving the user interface.