Systems and methods for prioritizing job candidates using a decision-tree forest algorithm

• US 8,818,910 B1
• Filed: 11/26/2013
• Issued: 08/26/2014
• Est. Priority Date: 11/26/2013
• Status: Active Grant

• Current Assignee: Comrise, Inc.
• Original Assignee: Comrise, Inc.
• Inventors: Liu, Sincia, Jiang, Fan, Bigini, Robert J.
• Primary Examiner(s): OUELLETTE, JONATHAN P
• Application Number:
  US14/090,225
• Time in Patent Office: 273 Days
• Field of Search: 705 11-912
• US Class Current: 705/321
• CPC Class Codes: G06Q 10/1053 Employment or hiring

Abstract

Computer-based systems and methods for prioritizing a list of job candidates for a job opening(s) with a hiring firm use the Random Forest algorithm, in which a number of decision trees are generated and the job candidates'"'"' respective probabilities of being the right fit for the hiring firm and for the position(s) are computed using the forest of decision trees. The job candidates may then be ranked according to their computed hiring probability. Non-conventional variables can be used, such as social media-related variables and external industry variables, such as the candidates'"'"' driving records. Other non-conventional variables that can be used include results from a personality or behavior assessment, results from a third-party skill assessment, and performance evaluations for prior projects or jobs of the candidates.

Claims

1. A system for prioritizing a list of job candidates for a job opening with a firm, the system comprising:

at least one computer system comprising at least one processor and at least one memory unit that is in communication with the at least one processor, wherein the at least one processor is programmed to;

use a Random Forest Algorithm to generate a quantity T decision trees, where T is greater than or equal to 2, and the T decision trees are trained based on a training data set of historical candidates classified into hired or unhired classes, wherein each decision tree of the T decision trees is generated by selecting historical candidates in the training dataset randomly with replacement to train the decision tree; and

for each of the job candidates, determine a classification of the job candidate for each of the T decision trees;

compute a hiring probability for each job candidate based on a ratio of (i) a total of number of times the job candidate was classified as hired to (ii) T; and

rank the job candidates by hiring probability to generate a prioritized list of the job candidates.

2. The system of claim 1, further comprising a remote computer system that sends the list of job candidate to the at least one computer system via a data network, and wherein the at least one processor of the at least one computer system is further programmed to transmit the prioritized list of the job candidates to the remote computer system.

3. The system of claim 1, wherein:

there are N historical candidates in the training data set; and

the at least one processor is programmed to select, for the training of each of the T decision trees, N historical candidates in the training dataset randomly with replacement.

4. The system of claim 3, wherein the at least one processor is programmed to generate the T decision trees by performing steps that comprise:

generating a N×

M matrix, where each of the N rows in the matrix corresponds to one of the historical candidates in the training dataset, and each of the M columns corresponds to an input variable for classifying the candidates;

to construct each of the T decision trees, at each node of each of the T decision trees, selecting randomly m out of M input variables; and

determining a best split on the m input variables to split the node.

5. The system of claim 4, wherein at least one of the input variables comprises a social media network variable.

6. The system of claim 5, wherein the social media network variable comprises a variable indicative of whether a candidate has a social media network account.

7. The system of claim 5, wherein the social media network variable comprises an online social media influencer score for the candidates.

8. The system of claim 4, wherein at least one of the input variables comprises a driving record variable for the candidates.

9. The system of claim 4, wherein at least one of the input variables comprises a result of a behavioral assessment.

10. A computer-based method for prioritizing a list of job candidates for a job opening with a firm, the method comprising:

using a Random Forest Algorithm to generate, by a computer system that comprises at least one processor and at least one memory unit, a quantity T decision trees, where T is greater than or equal to 2, and the T decision trees are trained based on a training data set of historical candidates classified hired or unhired classes, wherein each decision tree of the T decision trees is generated by selecting historical candidates in the training dataset randomly with replacement to train the decision tree; and

for each of the job candidates, determining, by the computer system, a classification of the job candidate for each of the T decision trees;

computing, by the computer system, a hiring probability for each job candidate based on a ratio of (i) a total of number of times the job candidate was classified as hired to (ii) T; and

ranking, by the computer system, the job candidates by hiring probability to generate a prioritized list of the job candidates.

11. The method of claim 10, further comprising:

receiving, by the computer system, the list of job candidate from a remote computer system via a data network; and

transmitting, by the computer system, the prioritized list of the job candidates to the remote computer system.

12. The method of claim 10, wherein:

there are N historical candidates in the training data set; and

selecting, for the training of each of the T decision trees, comprises selecting N historical candidates in the training dataset randomly with replacement.

13. The method of claim 12, wherein generating the T decision trees comprises

generating a N×

M matrix, where each of the N rows in the matrix corresponds to one of the historical candidates in the training dataset, and each of the M columns corresponds to an input variable for classifying the candidates;

to construct each of the T decision trees, at each node of each of the T decision trees, selecting randomly m out of M input variables; and

determining a best split on the m input variables to split the node.

14. The method of claim 13, wherein at least one of the input variables comprises a social media network variable.

15. The method of claim 14, wherein the social media network variable comprises a variable indicative of whether a candidate has a social media network account.

16. The method of claim 14, wherein the social media network variable comprises an online social media influencer score for the candidates.

17. The method of claim 13, wherein at least one of the input variables comprises a driving record variable for the candidates.

18. The method of claim 13, wherein at least one of the input variables comprises a result of a behavioral assessment.