KR102721044B1 - Method for culture fit matching job seeker - recruiter using machine learning - Google Patents

Abstract

A job seeker-employer culture-fit matching method according to an embodiment of the present invention is a job seeker-employer culture-fit matching method using machine learning performed by a computing device, and may include a step of clustering a plurality of job seekers into individual groups based on personal data such as the job seekers' careers and academic backgrounds; a step of deriving the job seekers' work tendencies by receiving response information to a job seeker's work tendencies questionnaire; a step of clustering a plurality of job seekers into a plurality of corporate groups based on corporate reputation data and corporate information; and a step of matching the clustered individual groups with the clustered corporate groups.

Description

{METHOD FOR CULTURE FIT MATCHING JOB SEEKER - RECRUITER USING MACHINE LEARNING}

The present invention relates to a job seeker-employer culture fit matching method using machine learning. More specifically, the present invention relates to a culture fit matching method that can match job seekers and employers using machine learning based on the job seeker's career, the quantitative indicators of the employer, as well as the job seeker's work tendencies and the company's reputation.

Korean Patent No. 2200334 discloses a job application document-based job competency analysis matching system. This includes a recruiter's human resource management server and a job competency analysis matching server.

The recruiting side's human resource management server creates a digital competency dictionary by referencing the current employee mentoring evaluation form for the job seeker's structured data, unstructured data, and job application document data, and performs scoring and tendency analysis for each competency analysis element for the newly collected job seeker's structured data and job application document data based on the digital competency dictionary.

The job competency analysis matching server performs scoring for each competency analysis element on newly collected job seekers' structured data, unstructured data, and job application document data based on the digital competency dictionary, and then conducts a tendency analysis using the results.

This job application document-based job competency analysis matching system collects individual job seeker unstructured data via big content distribution platforms including job seeker terminals, SNS, blogs, and internet networks, so that it can utilize not only the information of the job seeker's career and education, commonly referred to as 'specs', but also the potential competencies and tendencies of the individual job seeker for matching. Therefore, it can have the effect of increasing the reliability of the job seeker's competency analysis and increasing the matching probability for the recruitment data of the employer.

The embodiments of the present invention match job seekers with recruiters using personal data such as job seekers' specifications, career, education, and qualifications, work tendency diagnosis data, and job seekers' corporate reputation data and corporate information (structured data). This enables cultural fit that matches not only objective and rational structured data such as the abilities and capabilities of the job seekers and recruiters, but also the cultural tendencies of the two, thereby increasing the possibility that the job seekers will emotionally and culturally blend in with the recruiters during actual hiring.

A job seeker-employer culture-fit matching method according to an embodiment of the present invention is a job seeker-employer culture-fit matching method using machine learning performed by a computing device, and may include a step of clustering a plurality of job seekers into individual groups based on personal data such as the job seekers' careers and academic backgrounds; a step of deriving the job seekers' work tendencies by receiving response information to a job seeker's work tendencies questionnaire; a step of clustering a plurality of job seekers into a plurality of corporate groups based on corporate reputation data and corporate information; and a step of matching the clustered individual groups with the clustered corporate groups.

The above personal data may be structured data such as the job seeker's career, education, and qualifications.

The step of deriving the above work tendencies can be accomplished by utilizing the information obtained when job seekers answer a prepared questionnaire in advance, and calculating each job seeker's score for each item of the job seeker's individual proactivity, self-confidence, responsibility, thinking style, and communication based on this. The questionnaire can include information on preferred company type, company preference, organizational life, compensation, and work tendencies.

The above corporate reputation data is text data, which is unstructured data, and may be text evaluated by a recruiter, such as a headhunter, regarding the company and transmitted from the reputation information provision server to the computing device.

The above corporate information is transmitted from the corporate information server to the computing device, and the corporate information server may be the electronic disclosure system operated by the Financial Supervisory Service or the credit information provision platform provided by Korea Ratings Data.

The above corporate reputation data is text data, which is unstructured data that records an evaluation of a corporate culture, and the corporate information is structured data including a corporate outline and financial information, and the step of clustering a plurality of recruiters into a plurality of corporate groups based on the corporate reputation data and the corporate information includes: a step of clustering a plurality of recruiters into a plurality of corporate groups by performing natural language processing on the corporate reputation data; a step of clustering a plurality of recruiters into a plurality of corporate groups using the corporate information; a step of combining the labels of the clustering results using the corporate data of each company with the labels of the clustering results using the corporate information; a step of re-clustering companies based on the combined label data corresponding to each company; and the clustering may be by a machine learning method.

According to the present invention, the clustering result is used by including the job seeker's work tendency diagnosis data and the recruiter's corporate reputation data in the matching of individual groups and corporate groups, so that a result matching the corporate culture and individual tendencies can be obtained. Therefore, more precise matching is possible than with existing methods.

In addition, when clustering corporate groups, the results of clustering using corporate information and the results of clustering corporate reputation data are combined and clustered again, making it possible to cluster corporate groups that fully reflect the two factors above.

Figure 1 illustrates a job seeker-employer matching system according to one embodiment of the present invention.
Figure 2 is a flow chart schematically illustrating a job seeker-employer culture fit matching method using machine learning according to one embodiment of the present invention.
Figure 3 is a flow chart specifically explaining the job seeker clustering and work tendency derivation steps of Figure 2.
Figure 4 is a flow chart specifically explaining the enterprise group clustering step of Figure 2.
Figure 5 is a conceptual diagram explaining the individual group-corporate group matching step of Figure 2.
Figure 6 is an example of a questionnaire that job seekers answer when diagnosing individual work tendencies to derive work tendencies in Figure 2.
Figures 7 to 9 are examples of the results of the personal work tendency diagnosis.

Hereinafter, embodiments of the present invention will be described with reference to the attached drawings. The attached drawings illustrate exemplary forms of the present invention, which are provided only to explain the present invention in more detail, and the technical scope of the present invention is not limited thereby.

The present invention proposes a job seeker-employer culture-fit matching method using machine learning. FIG. 1 illustrates a job seeker-employer matching system according to an embodiment of the present invention, FIG. 2 is a flowchart schematically showing a job seeker-employer culture-fit matching method using machine learning according to an embodiment of the present invention, FIG. 3 is a flowchart specifically explaining the job seeker clustering and work tendency derivation steps of FIG. 2, FIG. 4 is a flowchart specifically explaining the corporate group clustering step of FIG. 2, FIG. 5 is a conceptual diagram explaining the individual group-corporate group matching step of FIG. 2, FIG. 6 is an example of a questionnaire answered by a job seeker during an individual work tendency diagnosis for deriving work tendencies of FIG. 2, and FIGS. 7 to 9 are examples of individual work tendency diagnosis results.

As illustrated in Fig. 1, a job seeker (1)-employer matching system according to one embodiment of the present invention includes a matching server (2). The matching server (2) can collect information on the job seeker (1) through a network.

The information of a job seeker (1) may include personal data (13) and work aptitude diagnostic data (11).

Personal data (13) can be input from a job seeker's (1) terminal connected through a network. Personal data (13) can include information such as career, education, and qualifications, which are expressed as so-called 'specs'. In addition, address, place of origin, and geographical information can also be included. When each job seeker (1) accesses the matching server (2) through a terminal and inputs and transmits his/her personal data (13) through a web page, the matching server (2) can receive it and store it in a database.

This personal data (13) is the data that has been traditionally used when matching job seekers (1) and recruiters. These data reflect the results of learning, experience, and training that job seekers (1) have accumulated to acquire the skills necessary for the job. However, even among people who have accumulated similar experiences, cultural factors such as personal tendencies, work tendencies, and personal visions can all be different. The traditional job seeker (1)-recruiter matching method has matched companies using only this personal data (13), ignoring cultural factors, or matching has been done through a separate offline interview process.

However, as the supply of job seekers decreases due to the aging population and the negotiation power of job seekers increases, the young job seekers (1), so-called 'MZ generation', are increasingly prone to quitting their jobs within a short period of time even if they are hired by companies that match them only based on their 'specs'. In addition, as the values of job seekers (1) diversify, the factors that job seekers (1) consider, which used to be mainly focused on monetary compensation called 'annual salary', are diversifying to various values such as personal growth potential, fair compensation, welfare system, horizontal culture, and work-life balance. Therefore, if these cultural factors are ignored and matching is performed, precise matching will not be achieved. In addition, it is not easy to completely determine the cultural fit between the recruiter and the job seeker (1) in an interview conducted over a limited period of time.

In order to perform more precise job seeker (1)-employer matching by including cultural elements in the matching, the job seeker (1)-employer matching method using machine learning according to the present invention uses work tendency diagnosis data (11).

Examples of work tendency diagnosis are specifically illustrated in FIGS. 6 to 9. As illustrated in FIG. 6, information obtained by a job seeker (1) answering a question about work tendencies can be utilized as work tendency diagnosis data (11). The job seeker (1) views a questionnaire on a webpage provided by the matching server (2) to the job seeker's (1) terminal, and the job seeker's (1) terminal transmits data generated as an answer thereto to the matching server (2), thereby enabling the work tendency diagnosis data (11) to be secured in the matching server (2). The matching server (2) can receive the work tendency diagnosis data (11) and store it in a database.

Additionally, the matching server (2) can receive corporate information (33) from the corporate information server (3) via the network.

Corporate information (33) is structured data, and examples thereof include corporate sales, net profit, size, number of employees, and industry. Corporate information servers (3) may include, for example, the electronic disclosure system operated by the Financial Supervisory Service, credit information provision platforms such as Korea Ratings Data, and other securities company servers, or a combination thereof. Matching servers (2) can request corporate information (33) from corporate information servers (3) via API and receive a response. If corporate information servers (3) do not provide API services, matching servers (2) can obtain information via web crawling techniques.

The matching server (2) can also collect corporate reputation data (31). The corporate reputation data (31) is unstructured data and can be in text form. The corporate reputation data (31) can be obtained from a corporate information server (3), which is a combination of each of, for example, a reputation information provision server where relevant parties such as headhunters and recruiters collect and write evaluations of the company, an online community where employees of the company comment on their own company, groupware, anonymous bulletin boards, etc. The matching server (2) can store the corporate reputation data (31) in a database.

As explained above, corporate reputation data (31) is a database of text data collected by each company by collecting sentences written by people who know the company well. Unlike corporate information (33), which is objective and rational structured data, this corporate reputation data (31) is subjective and emotional unstructured data. For example, the reputation of a certain company may include sentences such as 'This company is suffocating', 'It is a family-like company', and 'The CEO is passionate'. In other words, it will contain information on the company's relative strengths and weaknesses, the management's tendencies, corporate culture, and organizational culture.

The matching server (2) uses the above data, i.e., personal data (13), work tendency diagnosis data (11), corporate information (33), and corporate reputation data (31), to implement a job seeker (1)-recruiter matching method according to one embodiment of the present invention as illustrated in Fig. 2. That is, the job seeker's (1) information includes tendency information, and the recruiter's information includes reputation information, thereby enabling matching based on culture fit.

As illustrated in FIGS. 2 and 3, a job seeker-employer matching method according to one embodiment of the present invention includes a step (S1) of clustering and deriving work tendencies based on personal data (13, quantitative information) of a job seeker (1).

Here, as illustrated in Fig. 3, a step (S11) of performing clustering based on quantitative information of an individual, i.e., personal data (13), can be standardized in a set format and organized into a vector. Depending on the job seeker (1), personal data (13), i.e., academic background, career, qualifications, etc., will appear in various ways. Numerous personal data (13) stored in the database of the matching server (2) can be organized in a vector format and used to cluster job seekers (1) into several groups.

Clustering can be performed by machine learning, especially unsupervised learning. For example, using K-MEANS clustering, many job seekers can be divided into K groups with similar personal data (13). For example, they can be clustered into 27 groups. Through this process, many job seekers (1) can be assigned to one of several groups.

Next, as illustrated in FIGS. 3, 6 to 9, a step (S12) of calculating work tendencies using work tendency diagnosis data (11) may be performed in the matching server (2). Work tendencies may be diagnosed based on data (work tendency diagnosis data) that a job seeker answers to a questionnaire as illustrated in FIG. 6. The questionnaire may include content such as corporate preference (e.g., a company that is growing steadily, a company that is growing rapidly, a company that allows overseas work, a company that allows domestic work, a company that guarantees employment stability), organizational life (e.g., a company that guarantees work-life balance, vertical decision-making, horizontal decision-making), compensation (e.g., a company that is obedient to orders is good if it pays a high salary, a company that pays a high salary does not prefer a company if the commuting distance is long), work tendencies (I want to do challenging tasks, I want to do ordinary work), etc.

The work tendency can be calculated by calculating the data answered by the job seeker to the above questionnaire using a rule base, and calculating each job seeker's score for the five items of proactivity, self-confidence, responsibility, thinking style, and communication, as shown in Figures 7 to 9. In addition, preference data can also be calculated by classifying the company type preferred by each job seeker, such as large corporations, startups, global companies, and public companies.

The clustering results and work tendency diagnosis results produced in the above process (S11, S12) can be stored in a database (S13).

Next, as shown in FIG. 2 and FIG. 4, a step (S2) of clustering companies into groups based on corporate reputation data (31) and corporate information (33) can be performed by the matching server (2).

Step S2 may include a step (S21) of clustering corporate reputation data (31) by natural language processing, as illustrated in Fig. 4. Text related to corporate culture included in corporate reputation data (31) is extracted and similar companies are clustered into groups by using their frequency, etc.

As a step S22, a matching server (2) may include a step of clustering companies using corporate information (33). As described above, corporate information (33) includes structured data such as corporate size, corporate location, and sales. This corporate information (33) is used as machine learning data to cluster a number of similar companies.

In steps S23 and S24, the labels for the groups of individual companies obtained as a result of clustering in step S21 and the labels for the groups of individual companies obtained by clustering using corporate information (33) in step S22 are combined by company, and the combined data is clustered again to divide each company into multiple groups.

That is, the clustering results derived using the corporate reputation data (31) of each company and the clustering results derived using the corporate information (33) are combined, and clustering is performed again using this combined result. Through this process, the corporate reputation data (31) and the corporate information (33) can be combined to assign each company to one of multiple groups. By dividing the companies into groups so that both the corporate reputation data (31) and the corporate information (33) are reflected, it has become possible to include cultural factors in the matching of job seekers (1) and recruiters.

As shown in Fig. 2, when the individual group clustering step (S1) and the corporate group clustering step (S2) are completed, the corporate group (39) and the individual group (19) are matched (S3). As an example of the individual group (39)-corporate group (19) matching, a table (29) can be used as shown in Fig. 5.

In the table (29), multiple individual groups (19) are arranged horizontally and multiple corporate groups (39) are arranged vertically. A cell marked with 'O' means that a specific individual group (19) and a specific corporate group (39) are matched, and a cell marked with 'X' means that they are not matched. That is, since the cell values corresponding to individual group 1 and corporate group 1 in Fig. 5 are 'O', individual group 1 and corporate group 1 are matched.

One individual group (19) and one corporate group (39) are not matched one-to-one, but multiple individual groups (19) can be matched to multiple corporate groups (39). That is, as shown in the table of Fig. 5, corporate group 1 can be matched to individual group 1 and individual group 3. In addition, individual group 3 can be matched to corporate group 1 and corporate group 2. In the table (29), information on whether or not there is a match may be stored in binary form, or a real number indicating a matching rate may be stored. Therefore, for example, individual group 1 and corporate group 1 can have a matching rate of 10.2, and corporate group 1 and individual group 3 can have a matching rate of 98.6.

In Fig. 5, it is illustrated in a table format so that it can be understood by human readers, but the method of matching each individual group with each corporate group can be coded in various ways, and the illustrated table (29) is written to explain that groups can be matched rule-based according to set criteria.

The preset criteria for how to match individual groups (19) and corporate groups (39) can be provided in various ways. For example, if there is an expert or group of experts, such as a headhunter or recruiter, who has accumulated intuition about the real world and can know which individual groups are well suited to which corporate groups just by looking at the results of steps S1 and S2, the matching server (2) can receive information from the experts in advance and store it in a table so that their professional wisdom can be reflected, and then perform step S3. For this purpose, the matching server (2) will be equipped with appropriate input/output means.

Or, the satisfaction level of job seekers who have actually been employed or recruiters who have actually succeeded in recruiting may be surveyed and reflected to form a table (29). Or, a preferred corporate group or preferred individual group of a specific job seeker may be input, and this may be implemented for a number of recruiters or job seekers (1) to form a table (29) with this data. Machine learning may be used in the process of forming the table (29) and the matching process.

In step S1, clustering (S11) and work tendency derivation (S12) were performed as separate steps. In the matching step (S3) of the corporate group (39) and the individual group (19), the matching step (S3) may be performed using only the individual groups (19) distinguished by clustering (S11), or the matching step (S3) may be performed using the individual groups (19) divided by rule base according to the individual work tendency derived by work tendency derivation (S12), or the individual groups (19) derived by combining these may be used.

Alternatively, the individual groups (19) derived by the work tendency derivation (S12) in the first step are matched with the corporate groups (39), and only the individuals and companies matched in this matching are matched with the individual groups (19) derived by the clustering (S11) and the corporate groups (39) in the second step, thereby finally producing the matched individuals and companies.

Matching can be performed in various ways as described above, and accordingly, the table (39) should be prepared with separate criteria according to the matching method.

When matching (S3) is completed, a specific individual or multiple individuals within a specific individual group (19) that are matched to a specific company belonging to a specific corporate group (39) can be recommended (S4). Conversely, a recruiter suitable for the job seeker can be recommended. This can be done by the matching server (2) transmitting information to the terminal of the job seeker (1) or the terminal of the recruiter (not shown) through the network.

Those skilled in the art will recognize that the steps of FIGS. 2 to 4 may be modified in various ways. For example, the order of the steps may be modified in various ways, sub-steps may be performed in parallel, and the steps depicted may be omitted or other steps may be added.

Embodiments according to the present invention may be implemented in the form of program commands that can be executed through various computer means and recorded on a computer-readable recording medium. Examples of program commands include not only machine language codes created by a compiler, but also high-level language codes that can be executed by a computer using an interpreter, etc. A hardware device may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

Above, the embodiments of the present invention have been described, but it will be apparent to those skilled in the art that the present invention can be variously modified and changed by adding, changing, deleting or adding components, etc., within the scope that does not depart from the spirit of the present invention described in the claims, and this is also included within the scope of the rights of the present invention.

1; Job seeker 2; Matching server
3; Corporate information provision server; 11; Work tendency diagnosis data
13; Personal data 31; Corporate reputation data
33; Corporate Information

Claims (6)

A job seeker-employer culture fit matching method using machine learning performed by a computing device,
A step of clustering multiple job seekers into individual groups based on personal data such as job seekers' career and academic background;
A step for deriving the job seeker's work tendency by receiving response information to the job seeker's work tendency questionnaire;
A step of clustering multiple recruiters into multiple corporate groups based on corporate reputation data and corporate information;
A step of matching the above clustered individual group with the above clustered corporate group is included.
The above corporate reputation data is text data, which is unstructured data that records evaluations of corporate culture.
The above corporate information is structured data that includes the company's overview and financial details.
The step of clustering multiple recruiters into multiple corporate groups based on the above corporate reputation data and corporate information is as follows:
A step of clustering a large number of recruiters into multiple corporate groups by performing natural language processing on the above corporate reputation data;
A step of clustering multiple recruiters into multiple corporate groups using the above corporate information;
A step of combining the label of the result of clustering with the corporate reputation data of each company and the label of the result of clustering with the corporate information;
A step of re-clustering companies based on the combined label data corresponding to each company;
The above clustering is a job seeker-employer culture fit matching method using machine learning. In paragraph 1, the personal data is:
A job seeker-employer culture fit matching method that uses structured data such as the job seeker's career, education, and qualifications. In the first paragraph, the step of deriving the work tendency is,
This can be done by using the information obtained by job seekers answering a pre-prepared questionnaire, and calculating each job seeker's score on the items of individual job seeker's proactivity, self-confidence, responsibility, thinking style, and communication based on this.
The above questionnaire includes a job seeker-employer culture fit matching method that includes information on corporate preferences, organizational life, compensation, and work tendencies. In the first paragraph, the corporate reputation data is
It is text data, which is unstructured data.
A job seeker-recruiter culture fit matching method in which a text evaluating a company by a recruiter, such as a headhunter, is transmitted from a reputation information providing server to the computing device. In paragraph 1,
The above corporate information is transmitted from the corporate information server to the computing device.
The above corporate information server is a job seeker-employer culture fit matching method, which is an electronic disclosure system operated by the Financial Supervisory Service or a credit information provision platform provided by Korea Credit Data. delete