DonNTU   Masters' portal

Abstract

Content

Introduction

During the reception staff to the employer have to make important decisions that will continue to depend on the efficiency of the business and socio-psychological climate in the collective. When applying for a job, in addition to the skills of applicants for a particular position, they are important personal qualities: sense of duty, responsibility, punctuality, teamwork, friendliness and sociability. Wrong choice of professionals for employers with its negative consequences in the form of unnecessary material costs and lost work time.

Personnel selection – the process of managing a complex system with many objects which are, on the one hand, and the organization representing its employees, on the other – applicants for the position, having professional and personal skills. To automate the process of personnel selection is not possible to develop a universal model formalized because of the specific requirements for applicants from various organizations and institutions that act as employers in today's job market. At the same time, improve the management of this process can be by creating a decision support system for matching the level of specialist training to the labor market.

1. Theme urgency

Modern society is characterized by intensive development of techniques and technologies, as well as large volumes of processed information, which leads to a steady increase in the requirements for the training of specialists in various fields.

Today is a very urgent problem of finding skilled workers by employers to improve the efficiency of the enterprise. To solve this problem you need to select a limited number of "best" candidates for an interview with the employer [2]. Need to develop a system for the selection of candidates, which will be rated contenders. The initial data for the calculation of rankings - data profiles of applicants.

Who is most often used to evaluate the candidate interview. This method has several disadvantages. The perception of the candidate interviewer influence stereotypes, the first impression, physical attractiveness (unattractiveness), manner, location, clothing, and other factors, the key reason which can be called the subjectivity of the interviewer. Based on the above it seems necessary to develop a computerized system for the selection of personnel, with the passage of the testing phase, in which the estimate does not depend on the subjective factor and gives consistent results.

2. Goal and tasks of the research

The goal is to develop an intelligent system of selection of a limited number of applicants for a job interview with an employer.

The task of research is to optimize the selection process of a limited number of applicants for a vacant position for a job interview with an employer by using fuzzy logic methods.

Based on domain analysis can identify the following problems that must be solved system of selection of candidates:

  • Data preparation of profiles of candidates to a form suitable for further analysis.
  • Pre-selection profiles for advanced requirements and limiting factors.
  • Analysis of selected personal data.

    • 3. Review of Research and Development

    Analysis of the development of the problem of recruiting the staff showed that commercially available software products (Personnel Manager, «BOSS-Personnel", "Joy of the personnel", "1C: Payroll and Personnel Management 8.0", etc.) can not provide information analytical support to the decision on the selection of staff, with priority and contradictory to the demands of a staff that makes it necessary to develop a new method of computer personnel selection firm [ 2 ]

    As a software framework of a new method of selection of firm personnel are encouraged to use expert systems to support decision-making in crisis situations (ES PRKS) based on the automatic classification of the current situation in the face of incomplete and fuzzy input information and the formation of management solutions, corresponding to the detected class situation [ 3 ].

    Similar decision support system has already been developed by many companies, one of the used solutions for large enterprises are the SAP system [3]:

    Solution architecture SAPERPHCM

    Figure 1 – Solution architecture SAPERPHCM

    Since the module SAP E-Recruiting is a specialist in the recruitment of personnel, its use will involve the least amount of improvements that cover the functional deficits of the system. However, its use will be more effective in a joint operation with the component SAP Business Intelligence, using methods Data mining.

    Also well known are other ways to solve the problem of selection of candidates using neural networks [ 5 ] and decision trees [ 6 ]. The disadvantage of these methods is that they do not allow you to fully appreciate the professional quality of applicants in accordance with the requirements of employers. They make it possible to analyze only the data from the questionnaires applicants. However, the decision tree will be appropriate to use for the classification of skills, determination of the degree of importance and the dependencies between them.

    Known also a new approach to the assessment of applicants for jobs in the online recruitment system, using the taxonomy of machine learning algorithms to solve the problem of calculating the ranking of the candidate and the semantic matching techniques [ 7 ]. The proposed system extracts a set of objective criteria of the LinkedIn profile of the candidate, and compares them semantically with the requirements. It also displays them on the basis of personal characteristics of the linguistic analysis of their blog. The advantage of this system is that the results of its work is constantly compared with recruiters, and also that it can be used to automate the calculation of rating the candidates and the study of personality characteristics. Disadvantages: need for training selection algorithm to estimates of experts and a large labor input.

    The book Andreychikov A., Andreichikova O. "Analysis, synthesis, planning decisions in the economy" in Section 4.8.3, "Fuzzy inference in the problem by selecting a candidate for the vacant position of accountant," describes an approach that uses a fuzzy inference method for solving the problem of recruitment. The essence of which is as follows: the rules are compiled and stored in the knowledge base of intelligent system. In the process of solving the problem are defined by the user inputs that represent the values of the FIR linguistic variables corresponding to the requirements. The processing of these data is carried out by means of fuzzy inference procedures. The results of the system are multi fuzzy set obtained for a given candidate, and the measure of its similarity to the possible outcomes, ie fuzzy sets. As a result of sorting applicants were rating the candidates, who will select the best candidates who will go for a job interview [ 8 ].

    This method of decision-making using fuzzy inference rules is an adaptation of fuzzy logic to the decision-making processes with the original data in the form of point estimates. In some cases, evaluation of alternatives easier to produce using fuzzy numbers, which are the values of the linguistic variables. The rules can be applied not simultaneously but sequentially. Such an approach to computer support decision-making processes used in intelligent systems with fuzzy logic.

    advantage of this approach is that the formalization of knowledge by means of rules allows to take into account the importance of the different criteria and the rules themselves to form the requirements for the post. To account for the varying importance of the rules used in the book normalized weights that can be obtained either by pairwise comparisons, or by use of expert weights.

    In my work, I'll use the expert assignment of weights, as this will take into account the importance of each criterion and simplify the user experience with the system. In this problem, various possible approaches to the selection of the candidate for the post of: soft, hard, rational, and so I will use in their work tougher approach. Rigorous approach to the selection of the candidate for the position is possible in the case of an excess of qualified staff and resource time allocated for selection. The aim of this approach is to find the candidate that best matches the ideal.

    The disadvantage of this approach is that it does not take into account when evaluating the candidate of their professional data, not an algorithm for estimating the degree of possessing these skills are not described in the step of providing input to the system, as well as the task of recruiting the staff considered a narrow focus - only for the post Accountant.

    In my work I am I will consider these shortcomings and try to eliminate.

    Thus, to solve the problem of selection of candidates for the vacant post I will use the methods of fuzzy logic. The use of fuzzy logic to solve this problem will formalize the knowledge of experts in the formation of the requirements of the employer to the position of [ 9 ].

    4. Application of fuzzy parameters in the problem of the selection of personnel for the vacancy

    Give a formal statement of the problem of selection of personnel:

    1. Given a set of requirements Tm, imposed on the applicant for the position m:

    Tm = {P1m, P2m, ..., Pnm};

    Pim = {Qi1m, Qi2m, ..., Qikm};

    Pim – a set of values for each requirement;

    Qijm – value for the requirements for an applicant for a position m;

    i = 1, ..., n;

    j = 1, ..., k;

    m = 1, ..., M;

    n – number of requirements for a vacancy m;

    k – the number of values for each requirement;

    M – the number of vacancies;

    2. The set Gm factors – "the importance of the criteria" that determine the "level of importance" of all claims.

    Gm = {G1m, G2m, ..., Gnm};

    3. The set of characteristics Ls applicant s.

    Ls = {L1s, L2s, ..., Lzs};

    Lis = {Qi1s, Qi2s, ..., Qils};

    i = 1, ..., z;

    s = 1, ..., S;

    z – number of characteristics of the applicant s;

    l – the number of values of each attribute of the candidate;

    S – the number of applicants for the vacancy m;

    L - the set of all applicants.

    Requires:

    Select many applicants for the position P m – the required number of applicants in accordance with the requirements imposed on them.

    P = {P1, P2, ..., PS};

    One of the main challenges for the construction of the system of selection of candidates, is the problem of the definition of Gm – «criteria of importance».

    Selection of applicants for the vacancy m:

    1. Creating a list with the specified requirements.

    2. The definition of "criteria of importance" of each requirement by the expert use of weights.

    3. Determine the extent to which job applicants through a decision tree algorithm, and Sugeno fuzzy inference.

    4. Creating a list of "best" candidates.

    Assumes the use of these data as linguistic variables to form a ranking of applicants by fuzzy inference[10]:

    Type of activity (analyzed at the stage of selection of candidates from the database of the agency), date of birth (on the basis of which will be calculated age) general labor work experience, work experience in the vacant position, gender, education, occupation and qualifications, positions held, reasons for dismissal; experience of working in senior positions; computer experience, language skills, etc.

    Age (calculated based on date of birth) [11]:

    Figure 3 shows a view of the membership function for the age.

    Figure 3 – Graph of the membership function for the linguistic variable "Age"

    Figure 3 – Graph of the membership function for the linguistic variable "Age"

    Similarly, one can imagine other linguistic variables.

    Total work experience of work (in years):

    Experience with the vacancy (in years):

    Work experience in management positions:

    Gender:

    Education:

    Speciality:

    Qualification:

    Positions held:

    Reasons for dismissal:

    Computer experience:

    Language and other skills:

    The main parameters that are most often found in the demands of employers and have the greatest significance for them were selected by us. In the preparation of a list of demands to the employer must specify a value for each skill. This should be an opportunity to ask an unlimited number of requests, with the ability to edit the list of possible skills.

    Degree possession of skills can be obtained using a decision tree.After analyzing the requirements of Ukraine's largest site of search job and staff – work.ua, were chosen common skills for the construction of a decision tree of the Information Technology (IT). For the IT department decision tree will look like this:

    The Tree of solutions for the task staff recruitment in the IT department

    Figure 3 – The Tree of solutions for the task staff recruitment in the IT department
    (animation: 5 frames, 5 cycles of repetition, 20 kilobytes)

    Consider the data taken from the requirements of the vacant position, and the characteristics of the candidates of their profiles from the site work.ua.

    Vacancy number 1

    Developer. NET / C #

    24,000 USD., based on interviews

    Company:

    Ukrinvent, Ltd.

    All jobs from this company

    City:

    Kiev

    Employment Type:

    full-time

    Job Requirements

    • experience of 1 year

    • does not matter, are willing to take a student

    Job Description

    young companies to create applications for Windows requires adequate able to work in a team of developers. NET / C #.

    requirements.

    • Knowledge and experience of working with. NET Framework / C #

    • Knowledge of OOP principles

    • Skills in user interface design WPF

    • Network technology, WCF

    • Understanding the operating system architecture, and how it works.

    Conditions.

    We provide an office in the metro area Shuliavska, a friendly atmosphere, the presence of fellow programmers in the team, innovative and exciting challenges, loyalty management, a flexible schedule (possibly overlapping with studies).

    Salary worthy based on interviews and doing a bit of test task.


    We will make the rule base for the basic requirements of the questionnaire. For this we use the algorithm of fuzzy inference Sugeno. The advantage of this algorithm is that the construction of the rules it is possible to set the weights for each of the sub-conditions.

    Input variables:

    Let x1 – experience (work experience) of 1 year,

    x2 – knowledge and experience of working with. NET Framework,

    x3 – Knowledge and experience with C #,

    x4 – knowledge of OOP principles,

    x5 – skills in user interface design WPF

    x6 – Network Technologies, WCF.

    rule base for job number 1:

    If x1 is "any", the truth degree of sub-conclusion:

    W = 0,2 C1. Weight rules: F1 = 0,3.

    If x2 is "yes" and x3 is "yes", then the degree of truth sub-conclusion:

    W = 0,8 0,8 C2 C3. Weight rules: F1 = 0,6.

    If x4 is "yes", and x3 is a degree of truth-conclusion:

    W = 0,9 C4. Weight rules: F1 = 0,9.

    If x5 is "yes" and x6 is "yes", then the degree of truth sub-conclusion:

    W = C5 C6. Weight rules: F1 = 1.

    Consider the data of 2 profiles taken from the site work.ua. In order to preserve the confidentiality of information about the candidates, their personal data will not be analyzed, have been changed.


    Questionnaire 1

    Petrov Peter

    system administrator, from 2000 UAH. / mo., full-time

    Date of Birth:

    June 3, 1989 (23)

    City:

    Petrovsk

    Contact Information

    To view the contact information you need to log in as an employer or create an account.

    Experience

    administrator

    at 09.2011 to 01.2007 (4 years 8 months)

    next (Local Area Network)

    Responsibilities included subscriber connection, configure the client computer to the network, consultations (phone mode) for general issues concerning the work of LCS

    Education

    University

    with 09.2006 for 06.2011

    Kherson National Technical University, cybernetics, automated systems, Kherson.

    Skills

    Working knowledge of computer software

    znaenie OS Windows, Linux, MS Office, web-programming

    language

    Russian – expert

    Ukrainian – expert

    English – Intermediate

    More information

    purposeful, responsible, punctual, neat, new knowledge and skills come easy.


    Ñalculate the degree of truth for the sub-conditions of the first rule for the first sub-conditions. The triangular membership function is calculated as follows:

    X1 = 2 years

    For the first term = initial experience.

    B11 = (a - x) / (a-b), since experience "initial" [0, 3], b <= x <= c

    A, b, c - point for the triangular membership functions,

    B11 - degree of truth of the 1st sub-conditions for the 1st term.

    B11 = (3 - 2) / (3 - 1,5) = 0,67,

    For the second term experience = small.

    B12 = 0 because x <= a

    Similarly, for the remaining term: B13 = 0, B14 = 0, B15 = 0.

    Bi - the degree of truth for all the sub-conditions.

    B1 = max (Bi + j) = 0,67;

    M '(y) = average (Bi), to verify all claims is more suitable.

    M '(y) = 0,2 / 1 = 0,2;

    Ci – the degree of activation of c sub-conditions the weight of the rules.

    C1 = M '(y) * F1, where F1 – weight rules.

    C1 = 0, 2 * 0,3 = 0,06;

    Wi – sub-conclusion degree of truth.

    W1 = E1 * B1, where E1 – weighting factor in the rule base.

    W1 = 0,2 * 0,67 = 0,134 – sub-conclusion degree of truth for the 1st rule.

    Similarly, calculate the degree of truth sub-conclusion (Wi) and the degree of activation (Ci), we get:

    For the second rule:

    B1 = max (Vij) = 0.5, B2 = max (Vij) = 0;

    B1 – degree of truth. NET, which is determined according to the data of the decision tree.

    B2 – truth degree C #, is equal to 0, as does not possess this skill, according to the decision tree.

    M '(y) = average (Bi), to verify all claims is more suitable.

    M '(y) = 0,5 / 2 = 0,25;

    C2 = M '(y) * F2, where F2 – weight rules.

    C2 = 0,25 * 0,6 = 0,15;

    W2 = E1 * B1 + E2 * B2, where Ei – weighting factor in the rule base.

    W2 = 0,8 * 0,25 + 0,8 * 0 = 0.2 – truth degree sub-conclusion the 2nd rule.

    C3 = 0, W3 = 0, C4 = 0, W4 = 0, since the candidate has no such skill.

    Using the formula, the center of gravity for singletons get:

    y = sum (Ci * Wi) / sum (Ci);

    y = (0,2 * 0,134 + 0,15 * 0,2 +0 * 0 0 * 0) / (0.2 + 0.15 + 0 + 0) = 0.162 – the degree of compliance questionnaire number 1 Jobs number 1.


    Questionnaire 2

    Vasily Vasilyev

    C # developer, full time, part time

    Date of Birth:

    March 24, 1993 (20 years)

    City:

    Vasilyevskaya

    Contact Information

    To view the contact information you need to log in as an employer or or create an account.

    Education

    Some College

    with 09.2009 for 09.2014

    KPI FIOT, Kiev.

    More information

    Vasily Vasilyev

    Locations: Vasilyevskaya, st. Vasilevsky, 18/20

    Telephone: [view contacts]

    E-mail: [view contacts]

    Objective: Improvement and development of practical skills in a real-world projects.

    Education:

    2009 – present – Kiev Polytechnic Institute, Department of Computer Science, specialization: "Systems Engineering»

    experience and work:

    • Experience in the development of commercial applications do not have

    Knowledge and Skills:

    • knowledge of C #,. NET Framework;

    • basic knowledge of Flex / Flash, ActionScript;

    • understanding of OOP principles;

    • professional PC user;

    • information systems, relational databases;

    • Technical knowledge of English;

    Personality:

    Ability to work well. Honesty in work. Hard work.

    Interested to improve and develop new technologies and development tools.

    Wishes:

    • flexible working hours.


    calculate the degree of truth for the sub-conditions of the first rule for the first sub-conditions.

    X1 = 0;

    For the first term = initial experience.

    B11 = 0 because x <= a

    Similarly, for the remaining term: B12 = 0, B13 = 0, B14 = 0, B15 = 0.

    Bi – the degree of truth for all the sub-conditions.

    B1 = max (Vij) = 0;

    M '(y) = average (Bi), to verify all claims is more suitable.

    M '(y) = 0/1 = 0;

    Ci – the degree of activation of c sub-conditions the weight of the rules.

    C1 = M '(y) * F1, where F1 – weight rules.

    C1 = 0 * 0.3 = 0;

    Wi – sub-conclusion degree of truth.

    W1 = E1 * B1, where E1 – weighting factor in the rule base.

    W1 = 0,2 * 0 = 0 – sub-conclusion degree of truth for the 1st rule.

    Similarly, calculate the degree of truth sub-conclusion (Wi) and the degree of activation (Ci), we get:

    B1 = max (Vij) = 1, B2 = max (Vij) = 1;

    M '(y) = average (Bi), to verify all claims is more suitable.

    M '(y) = 2/2 = 1;

    C2 = M '(y) * F2, where F2 – weight rules.

    C2 = 1 * 0,6 = 0,6;

    W2 = E1 * B1 + E2 * B2, where Ei – weighting factor in the rule base.

    W2 = 0,8 * 0,8 * 1 + 1 = 1.6 – truth degree sub-conclusion the 2nd rule.

    For the third rule:

    Bi – the degree of truth for all the sub-conditions.

    B1 = max (Vij) = 1;

    M '(y) = average (Bi), to verify all claims is more suitable.

    M '(y) = 1/1 = 1;

    Ci - the degree of activation of c sub-conditions the weight of the rules.

    C3 = M '(y) * F3, where F3 – weight rules.

    C3 = 1 * 0,9 = 0,9;

    Wi - sub-conclusion degree of truth.

    W3 = E1 * B1, where E1 – weighting factor in the rule base.

    W3 = 0,9 * 1 = 0.9 - sub-conclusion degree of truth for the 3rd rule.

    For the fourth rule:

    Bi – the degree of truth for all the sub-conditions.

    B1 – truth degree of WPF, which is determined according to the data of the decision tree.

    B2 – the degree of truth of WCF, which is determined according to the data of the decision tree.

    B1 = max (Vij) = 0,5;

    B2 = max (Vij) = 0,5;

    M '(y) = average (Bi), to verify all claims is more suitable.

    M '(y) = (0,5 +0,5) / 2 = 0,5;

    Ci – the degree of activation of c sub-conditions the weight of the rules.

    C4 = M '(y) * F4, where F4 – weight rules.

    C4 = 0,5 * 1 = 0,5;

    Wi – sub-conclusion degree of truth.

    W4 = E1 * B1 + E2 * B2, where E1, E2, – weighting coefficients in the rule base.

    W4 = 1 * 1 + 0.5 * 0.5 = 1 – degree of truth sub-conclusion for the 4th rule.

    Using the formula, the center of gravity for singletons get:

    y = sum (Ci * Wi) / sum (Ci);

    y = (0 * 0 * 0.6 1.6 + 0.9 * 0.9 0.5 * 1) / (0 +0.6 +0.9 +0.5) = 1.135 – the degree of compliance questionnaire number 2 Job number 1.

    Conclusion: 1.135> 0.162, therefore, the candidate with the application form number 2 is more appropriate then the candidate with the application form number 1.

    Based on the rule base can be generated a ranking's table of applicants, with the help of fuzzy methods of inference. To improve the quality of the system is planned to add special parameters, according to the vacant post.

    Conclusion

    Qualitative data analysis is possible if the complete information of personal data of candidates. Thus, the process of assessing the compliance level of training applicants advisable to carry out the requirements of the employer with the use of an intelligent systemsystem, the work of which will provide an objective assessment of compliance with the level of training of applicants requirements of the employer.

    This paper describes the task of creating a list of applicants for pre-selection of candidates by using fuzzy logic. The use of fuzzy indicators will take into account all the requirements of employers, and select the best candidates to interview.

    This master's work is not completed yet. Final completion: December 2013. The full text of the work and materials on the topic can be obtained from the author or his head after this date.

    References

    1. Gil M., Fonotov A., Development of a preliminary selection of candidates for interview on the basis of data mining techniques, scientific and technical journal SCC of Ukraine "Artificial Intelligence» ¹ 2 – 2013, pp. 24–34.
    2. Kretov V., Korobitsyn I., Computerized method of selection of the personnel of the company / / Science of Science online journal [Electronic resource]. – Access mode:http://naukovedenie.ru/index.php?id=160.
    3. Kretov V., Korobitsyn I., A new method for automatic classification of acts of terrorism / / Scientific and technical information. 2006. Number 5. pp. 14–16.
    4. SAP Business Management Software Solutions, Applications and Services [Electronic resource]. – Access mode: http://www.sap.com/.
    5. Azarnova T. Ternovykh I., Application of neural networks in the recruitment process, and telling Series: System Analysis and Information Technologies, 2009. ¹ 2. pp. 76–80.
    6. Qasem A. Al-Radaideh, Eman Al Nagi, Using Data Mining Techniques to Build a Classification Model for Predicting Employees Performance, (IJACSA) International Journal of Advanced Computer Science and Applications, ¹ 3, ¹ 2, 2012, ñ. 144-151.
    7. Faliagka E. et. al., Taxonomy Development and Its Impact on a Self-learning e-Recruitment System, International Federation for Information Processing, 2012 pp. 164-174.
    8. Andreychikov AV Analysis, design, planning decisions in the economy. / A. Andreychikov, ON Andreichikova. - Moscow: Finances and Statistics, 2000. — 203 pp.
    9. F. Herrera, E. Lopez, C. Mendana and M.A., Rodriguez, A linguistic decision model for personnel management solved with a linguistic biobjective genetic algorithm, Fuzzy Sets and Systems, 2001, ¹ 118, Issue 1, pp. 47-64.
    10. Gil M. V., Fonotov A. M, Data mining as part of the information technologies market of the human resources, Information control systems and computer monitoring (ICS and CM) // Materials III National Conference of Students and Young Scientists. - Donetsk, Donetsk National Technical University – 2012, ðð. 157-160.
    11. Gil M., Fonotov A., Personnel selection procedure for a position based on fuzzy parameters, Information control systems and computer monitors (ICS 2013) / Proceedings of the IV International scientific conference of students and young scientists. - Donetsk, Donetsk National Technical University - 2013, Volume 2, p. 67-71.