The specific features of the tasks of managing skilled human resources during recruiting, which allow to identify them as a task of multicriteria analysis and decision making in a fuzzy environment are highlighted in the article. A generalized conceptual model of decision-making in the recruitment tasks of IT specialists based on collected and analyzed data from social networks is proposed. It is substantiated that in order to increase the efficiency and transparency of solutions during recruiting, it is expedient to use multi-criteria optimization based on the TOPSIS method and show its advantages. The modification of the TOPSIS algorithm for recommending recruitment by qualified human resources on the basis of the analysis of user profiles in social networks for the Big Five model was proposed. The modification consists in integrating additional constituent management components into the decision-making algorithm, which provides calculation based on the analytical hierarchy process (AHP) criteria of the psychological state of IT specialists. Using the methods TOPSIS, AHP and scales for assessing the psychological state of skilled human resources, an example of recruitment has been used to make experimental calculations on the ranking of applicants, which demonstrated the efficacy of the proposed approach.
Today, professional recruitment is a necessary and necessary process in every
organization in any sphere of activity: its financial performance depends on the quality and
efficiency of the given process [
This allows us to find professionals of a certain level for the corresponding position. But it does not guarantee solving problems, compliance with the human factor for the fulfillment of the respective official duties and the effectiveness of the work of the respective claimant. As well as the acquired professional knowledge and the experienced experience on the quality of work affects the psychological portrait of the very IT specialist. One of the modern tasks of analyzing the psychological state of personality when recruiting IT specialists is to analyze his profile in social networks (posts on forums, comments on events, activity in social networks, etc.). The most popular social networks for IT professionals are LinkedIn, Facebook, Twitter, Instagram, Google+, Youtube. Analysis of the level of activity in relevant social networks of relevant IT specialists and NLP analysis of the author's and content content of these individuals will enable to automatically form the profile of the psychological state of the applicant for the elected post. Ability to choose from a plurality of applicants, in addition to the level of competencies and even the psychological factors without subjectivity from the recruiter, will greatly improve the efficiency of recruitment for companies. Qualitative recruitment will not only save them time and resources, but will allow them to increase their own profits. The purpose of this study is to develop a technology for making managerial decisions in recruiting tasks in accordance with the above conceptual model and in verifying its effectiveness. 2
Influence of social networks for formation of psychological
portrait of the person
Social networking service is a web site or other Internet service that allows users to
create a public or semi-public questionnaire, make a list of users with whom they
have a connection and view their own links list and lists of other users. The nature and
nomenclature of connections may vary depending on the system. Unlike social
networking services, the user is not in the center of the Internet community; the user's
relation to other members of the community is in the foreground. The focus of the
Internet community is on the user's contribution to achieving common goals, values
and communication. In social networks, the user is in the center of the system and can
belong to more than one group at a time [
Secondary connections in social networks are relatively weaker than secondary
connections in Internet communities. For example, if you have several social links in the
social network User2 , connections with unknown to him User2 will be weaker than
similar connections in the Internet community. In addition to providing useful
services, social networking Internet services are also a number of dangers [
Analysis of social networks has a large potential for use in information operations,
since it allows us to investigate the attitude, outlook and communication of a wide
range of individuals, in particular [
Accordingly, content analysis can be used to find individuals in the process of
recruiting (Fig. 1). Geocoded posts can complement the analysis, and help to assess the level
of activity of such personality civility. Due to the analysis, it is possible to expand the
analysis of the applicant for a post. The simpler classification of IT groups is
sufficiently well describes the main lines of work at the lower (production) level of the IT
company, that is, they are practically in all companies (test manager, project manager,
business analyst, sysadmin, designer, team lead, architect, programmer, IT sales).
Each group must possess not only relevant professional qualifications and experience,
but also psychological features. According to the European e-competence Framework,
the list of IT specialties forms 23 specialties, in particular [
Candidates
Actions
Request Support service
Request Recruiting system Rules Request
Content Report
Social
Networks
The company
The list of professions somehow not got the following, which are even in Ukraine
[
writer), 8. Top Manager.
4. Designer,
Although the hierarchy recommended in the main IT professions is the most
commonly used in Ukraine [
Candidates
Action Action
Social Networks
Rules
Recommended
Not Recommended
Model Big Five for determination of the psychological state of
personality
Regardless of its professionalism and level of experience, the Big Five model is used
to determine the main features of a psychological portrait of a person. But it is usually
used with the testing of these personalities. And on the sow of their answers build a
corresponding psychological portrait. The modified version of the definition based on
the Big Five model is the proposed method by the Chinese Bai Shuotian [
Candidates
Actions
Recommended
List
List
Results
Rules Results
Social Networks
Parser Recruiters
Analitics Report
Not
Recommended
Report Requests
Company
1. Periodic generalization of requirements for a position to formulate
recommendations [
Model Big Five is a hierarchical model of personality, which describes the five
main features that make up the personality of a person. For the best memorization in
the English literature, five rice are in the acronym OCEAN [
Since the interpretation of results depends on the natural language that a person thinks, many countries of the world have their own versions of testing. The model is based on variables that are most widely spoken in the natural language through the use of a certain emotional color in the lexicon of the corresponding adjectives, nouns and rotations, as well as the frequency of their use. At the same time, the assumption is assumed that the natural linguistic person is formed:
M BF C Ext ,C Agr ,CCns ,C Nrt ,COps , Ext, Agr,Cns, Nrt,Ops , (1) where the components of the tuple: extraversion (engagement): sociability, assertiveness or rest, passivity, that is С Exp Exp(C,U Exp ) through parameters U Exp ( u1Exp is sociability / sociability through the frequency and content of emoticons and liches at posts and friends' messages, uExp is perseverance / rest due to the frequency of reactions to the 2 actions of friends, u3Exp is activity / passivity due to the frequency of activity in the profile); benevolence (pleasure): kindness, trust, warmth or hostility, selfishness, distrust, that is С Agr Agr(C,U Agr ) through parameters U Agr ( u1Agr is kindness / hostility through reaction to new events, establishing new contacts and answering new queries. u2Agr is trust / distrust due to the response to the invitation, u3Agr is warmth / egoism through frequent posts about oneself in relation to posts about social events); conscientiousness (reliability): organization, thoroughness, reliability or carelessness, negligence, unreliability, that is СCns Cns(C,U Cns ) through parameters U Cns ( u1Cns is organization / negligence due to the quality of author's posts, the degree of their chaos, u2Cns is solidity / serenity through the richness author posts, uCns is reliability / unreliability due to the reaction of friends to the author's 3 events, the frequency of friends' greetings with the events, the frequency of answers to requests and the posts of friends); emotional stability: relaxation, balance, resistance or neuroticism is nervousness, depression, irritability, neuroticism, that is С Nrt Nrt(C,U Nrt ) through parameters U Nrt ( u1Nrt is relaxation / nervousness through the frequency and content of emoticons, nicknames and posts to friends' posts, u2Nrt is Balance / depression due to content content, emoticons, and likes, u Nrt is stability / 3 irritability due to the frequency of responses); culture, openness to experience: spontaneity, creativity or limited, intermediate, narrow interests, that is СOps Ops(C,U Ops ) , where C is content from user social networking profiles, U Ops is rules for calculating the criterion, that is. U Ops {u1Ops ,u2Ops ,u3Ops ,u4Ops} ( u1Ops is the frequency of occurrence of words related to kindness/anger; u2Ops is frequency of occurrence of words related to trust / distrust, u3Ops is the frequency of occurrence of words related to heat / hostility, uOps is the frequency of occurrence of words related to sincerity / egoism).
4 The following parameters are the main indicators for forming the criteria for social user behavior according to the Big Five model: frequency and content of emotional coloration of likes for a certain period of time; frequency and content of emotional coloration of emoticons for a certain period of time; frequency and content of emotional coloring of posts; the frequency and frequency of activities in the social network; average period of stay in social networks, taking into account the period (working time, non-working hours, season, etc.); participation and activity in groups; number of friends; frequency of greetings of friends; the presence / absence of pages, their number and frequency of updates; presence / absence of author's content (photos, videos, author's posts, etc.), its volumes and frequency / time of appearance; the presence / absence of information about yourself, the level of its completeness; the frequency and content of the vocabulary and the broadcasts of the applicant / author's / general content; the frequency and content of the comments of friends on the broad content of the applicant; proportionality of accessible public content to the closed view for all visitors. This list can be continued, depending on the possibility of the social network itself. So in LinkedIn it is possible to further analyze the frequency of recommendations on the skills from friends (the level of trust of the surrounding to the applicant). Factors of the behavior model Big Five in research denote differently, but the overall content of the model is fairly stable and based on the following postulates []: all adult persons can be characterized by a specific combination of personality traits, influencing thoughts, feelings and behavior (on individuality); characteristics of the person being studied are endogenous basic tendencies (of origin); the features develop in the childhood, finally formed in adulthood and retain their immutability in the adapted subjects (on development); features are organized hierarchically, from narrow and specific to wide, generalized dispositions (about structure).
The described models are close to the theory of individuality. Theories of personality traits are intermediate between typological and idiographic (clinical) approaches to the study of individuality. However, for its use it is difficult to determine the correlation between different characteristics without the introduction of vertical and horizontal measurements, which are the basis of the hierarchy within the system of personality. Interpretation of rice as a situational sustained appearance also raises doubts. However, this does not forbid the use of the possibility of isolating and predicting personality traits. Today, the well-known corporation has developed an algorithm that defines features of the personality in English (Spanish) authoring text of 100 words or more. IBM Watson Personality Insights attempts to apply a linguistic analysis to determine the psychological portrait of its author along a piece of text [74]. The service itself is paid, it is offered for analysis of clients by blogs, tweets, and entries in the forum. But he has a demo where you can download a text of just one hundred words (it can be more, but one is the minimum). And the car, like a real oracle, will give a deployed personality trait. True, the text should be in English or Spanish [74]. To get a psychological portrait, simply insert a piece of any text written by the person in question in the appropriate field, and then click on the "Analyze" button. The program with a small delay gives a brief description, and when scrolling down the page you will find more detailed data. The text is analyzed for the presence of certain marked words, and the results are ordered by recognized psychological metrics, such as the Big Five - the assessment of personality on five characteristics, such as consciousness, friendliness, extraversion, emotionality and openness of experience. The service has APIs, open access documentation, and also a section on GitHub.
According to [74] the results of the analysis of the text of Richard Branson under the heading "My idea of paradise" showed that the entrepreneur prefers imaginations rather than facts, believes in the best in people, trusts them, and in many of his actions is very independent. These data, if considered relevant, can be useful not only for analyzing the client base, but also for HR-specialists in evaluating applicants. However, the service users evaluate ambiguously. Quora has a related topic, where many participants believe that the service of accuracy of characteristics can be compared to horoscopes, they say, in any collection of very general phrases, everyone can see themselves. But subjective thought. Without a large number of statistical data, with close cooperation with psychologists, one can not fully confirm the effectiveness, efficiency and quality of the results. In parallel, with modern recruiting, the recruits themselves do not have a psychologist's education, and their subjectivity significantly influences the outcome of the formation of proposals for positions among applicants. It also does not solve the problem of qualitative selection of personnel. Therefore, companies often suffer losses and lose time searching for relevant professionals in the current fast-paced time and growing rates of IT business development. But a couple of drawbacks to using the method. The main thing in our opinion is the disadvantage of such a system - it is possible to pick up in advance or write a text that will be interpreted as beneficial to the applicant. Nowadays there is only one source of an independent, unique collection of content that has been collected over a long period of time by a specific personality, the user of social networks. As a person has not tried, she can not play a role for a long time (years), not showing himself in relationships with people around him in the same social network. The history of the user profile of a social network user is an objective information about his psychological state based on the Big Five model. Only appropriate methods of data extraction and NLP should be used to form the set of criteria for its psychological portrait for recruiting. Another disadvantage is that when creating requirements for the positions of IT professionals, the company usually uses other criteria such as communicative, responsible, leadership qualities, teamwork and creativity. This list may be expanded, but it is usually a fundamental indicator. By collaborating with representatives of an IT cluster with a specialist in psychology, this problem can be easily solved using, for example, a structured technique of the analytical hierarchy process (AHP, decision making in a multi-arterial setting) as shown in Fig. 4. We will use the web resource http://victana.lviv.ua/matrytsia (Fig. 5) for a conditional example in the absence of statistical research in this area and active close cooperation on these issues with the IT cluster and psychologists using approximate calculations (Fig. 6). In fig. 7. and Table 1 shows the results of calculations of the criteria for our conventional case.
Position Communicativeness
Responsibility
Leadership
Team work
Creativity Openness
Conscientiousness
Extraversion
Agreeableness
Neuroticism
The process of determining the suitability of a person's psychological state for an IT specialty On the basis of a comprehensive approach to the accounting of the specifics of the processes of human resources management, a generalized conceptual model of decision-making in recruiting tasks will be provided by the following set of information: A set of feasible alternatives between applicants for a post
Cnd {cnd1, cnd 2 , , cnd n } {cndi , i 1, n} ; A set of criteria for choosing Big Five that characterize the psychological state of applicants / candidates (alternatives) – U {U Ext ,U Agr ,U Cns,U Nrt,U Ops} or U {U1, U 2 , , U m} {U j , j 1, m} , where m=5, since there are only 5 model criteria Big Five; A set of subcriteria characterizing each of the criteria –
U j {u j1, u j2 ,..., u jT } {u jt , t 1,T} ; The area of determining the values of each individual criterion – Vcr; A set of social networks where information about the applicant for the decisionmaking procedure is collected and analyzed – Ssn; A set of relationships between profiles in various Social Networks – Rps; The relationship between sets Cnd, U та Ssn – Rss; Linguistic expressions reflecting the degree of satisfaction of applicants to private criteria (degree of affiliation) – Lrd. Relations between criteria and private criteria – Rcp.
In order to achieve the goal as structured technique AHP, TOPSIS (The Technique for Order Preference by the Similarity to the Ideal Solution) has been selected in recruiting tasks, which allows to eliminate a number of disadvantages of existing instrumental approaches. The method is modified to the conditions of the accepted conceptual model of decision-making in recruiting tasks [75-80]. The main idea of the TOPSIS method is that the best bidder should have not only the greatest proximity to the ideal solution, but also beyond all other contenders for an unacceptable solution. Here, the best (optimal) solution is a vector that contains the maximum values for each criterion for all applicants, and the unacceptable (worst) solution is the vector containing the minimum values for each criterion. As follows from the essence of the TOPSIS method, using the latter can quite effectively solve the problem of fuzzy multicriteria optimization, which make up the mathematical basis of decision support in human resource management tasks. Under the multicriteria optimization in the decision-making theory, we mean the choice of the best solution among potential applicants. The TOPSIS method is one of the effective tools for promoting recruiting and experts in formulating their goals and subjective advantages, structuring the set of criteria, evaluating applicants in the decision making process in fuzzy mathematics, linguistic variables, fuzzy sets and fuzzy numbers. The solution of the optimization problem using TOPSIS involves the need to translate the values of qualitative linguistic variables, expressing the degree of satisfaction of one or another applicant to the criteria, in fuzzy numbers [75]. The fuzzy number is a fuzzy subset of a universal set of real numbers having a normal and convex membership function for which there is a carrier value, where the membership function is equal to one, while the function of membership decreases when the maximum or left moves from its maximum. According to [75] the fuzzy opinions of experts, for example, psychologists, formulated in terms of natural language, or independent indicators, calculated on the basis of analysis of social networks. These indicators can be described by fuzzy triangular and fuzzy trapezoidal numbers. In this paper, given the need to ensure the stability of the criteria to the boundaries of the interval of validity, an obscure trapezoidal number is used. Formation of judgments of the expert as a result of the analysis of the applicant's activity in social networks in the form of a fuzzy trapezoidal number in practice is realized in this way. The investigated object according to the selected criterion is evaluated by the expert by a four-digit number h (h1, h2 , h3, h4 ) , where hi are real numbers. The essence of this procedure is that the value of the criterion is in the range from h1 to h4 , but most likely it is within the range from h2 to h3 . If in the four the average numbers will be equal, that is h2 h3 , then a fuzzy trapezoidal number turns into an unclear triangular number. Using operations on membership functions based on the segment principle, operations on fuzzy numbers [75] are introduced. When using TOPSIS, some operations on fuzzy numbers should be taken into account. Let two fuzzy trapezoidal numbers be set as h (h1, h2 , h3, h4 ) and g (g1, g2 , g3, g4 ) . Below are the operations of summation, the difference and the product of these numbers: h g h1 g1, h2 g2 , h3 g3, h4 g4 , h g h1 g4 , h2 g3, h3 g2 , h4 g1 h g h1g1, h2 g2 , h3g3, h4 g4 h f h1 f , h2 f , h3 f , h4 f The distance between two fuzzy trapezoidal numbers is determined from the expression [75]:
If h = g , that is h and g equivalent, then wo (h, g) 0 . To implement the method, it is necessary to operate with linguistic variables and their values, expressing verbal scales for measuring signs. At the same time, the levels are arranged in order of increasing intensity of the manifestation of these features. In this case, the number of values (gradations) of the linguistic variables is seven. Table 2 shows the 7-level values of the linguistic variable and their corresponding fuzzy trapezoidal numbers. According to Table 2, numeric matching can be found for each value of the linguistic variable. For example, the numerical correspondence of the linguistic value "partially (2) (3) good", which is one of the gradation of the measurement of properties, is determined by the 100-point rating system as (5, 6, 7, 8). We present an algorithm for multicriteria optimization of recruiting tasks based on the TOPSIS method and the results of the analysis of the applicants' activity. Cnd {cndi , i 1, n} in social networks Ssn {sl , l 1, v} . Let the following components of recruiting tasks be known:
U {u j , j 1, m} ); criteria ( u j {u jt , t 1, T j } ); U j {u jt , t 1, T j } is a set of private criteria as sub-criteria of the Big Five model Ssn {sl , l 1, v} is a set of social networks; rcp j , j 1, m are the coefficients of the relative importance of the criteria ( rcp jt , t 1, T , j 1, m are the coefficients of the relative importance of private
rpsl , l 1, v is coefficients of reliability of social networks.
The purpose of the task is to rank the applicants on the basis of activity ratings in social networks, taking into account the reliability of the latter. The solution to the problem involves the following sequence of actions: Step 1. To conduct multi-criteria optimization of recruiting tasks based on the TOPSIS method, it is necessary first of all to get rid of the hierarchical structuring of the criteria. To this end, based on the AHP Saati method, we use the coefficients of the relative importance of the criteria and the private criteria to determine the weights [75], which will be taken into account by the integral criterion U {U j , j 1, m} . In a formalized form, the product rcp j , where m rcp j 1 and j1
Tj rcp jt 1 , where t1 is determined rcp Ujt is the weight of the private criterion u jt in the calculation of the integral criterion U {u j , j 1, m} , that is rcp Ujt rcp jt rcp j . As a result, the twolevel hierarchical structure of the selection criteria U {U j , j 1, m} , characterizing the applicant for a position, is reduced to a one-step hierarchy. In the next steps for simplification of the indexes, all private criteria are combined in a single set S. S {u jt , j 1, m, t 1, T j } {uz , z 1, Z} , z T j1 t , j 1, m, t 1,T j , T0 0 .(4) Here Z is the total number of private criteria that characterize the applicant for a m position, that is Z T j . In this case, rcp z rcp Ujt .
j1 Step 2. Degrees of membership (matching) of applicants to private criteria are estimated linguistic values and are expressed by fuzzy trapezoidal numbers Trnl (wilz ) (eilz , fizl , gilz , hilz ) . So, for example, if the degree of satisfaction (affiliation) of the applicant cndi private criterion uz on the basis of analysis of the applicant's activity in the social network l estimated value "good", then it is expressed as
wilz (7, 8, 8, 9) , and if this assessment is "very good", then wilz (8, 9, 10, 10) etc. As a result of the Big Five assessing the degree of membership of applicants for a job by private criteria, we obtain the following matrix:
Trnl [wilz ], l 1, q {eilz , fizl , gilz , hilz}, l 1, q .
Step 3. This step involves preliminary calculation of the reliability ratios of social networks rpsl , l 1, q . To this end, a modification of the method was introduced, which involves integrating into an algorithm an additional step, which involves calculating and introducing the reliability coefficients of social networks involved in the bidding process. From given the reliability coefficients of social networks rpsl , l 1, q matrix is formed
Trn rpsl [wirzpsl ], l 1, q {eirzpsl , fizrpsl , girzpsl , hirzpsl }, l 1, q .
Elements of this matrix are trapezoidal numbers that express the degree of satisfaction of the applicant cnd i private criteria uz taking into account the reliability of social networks and calculated as follows: eirzpsl eilz rpsl ; f rpst fizl rpsl ; girzpsl gilz rpsl ; hirzpsl hilz rpsl .
iz Step 4. The single matrix is determined:
Trn rpsl [wirzpsl ], l 1, q {eirzpsl , fizrpsl , girzpsl , hirzpsl }, l 1, q Trniz [wiz ] {eiz , f iz, giz , hiz } .
The elements of this matrix are defined as follows: eiz {min eirzpsl , l 1, q}; f iz 1 q f rpsl ; q l1 iz giz 1 q girzpsl ; hiz {max hirzpsl , l 1, q}. q l1 (5) (6) (7) (8) (9) Step 5. Elements of the matrix Trniz [wiz ] {eiz , f iz, giz , hiz } multiply by the weight of the private criteria. As a result of this operation a weighted fuzzy matrix is constructed Trnirzcp [wirzcp ] {eirzcp , fizrcp , girzcp , hirzcp } here: ercp eiz rcpz ; iz f rcp fiz rcpz ; iz girzcp giz rcpz ; hrcp hiz rcpz .
iz (10) Step 6. The resulting matrix is normalized. This method Hsu and Cehn is used [],on the basis of which are determined hz max hirzcp , i 1, n . Next on the basis of the expression:
TrniNz wiNz eiNz , fizN , giNz , hiNz eirzcp , hz fizrcp , girzcp , hirzcp hz hz hz .
The elements of a normalized decision matrix are determined.
Step 7. Based on the weighted meanings, the ideal positive (best) solution is determined (IPS) Cnd * . To this end, for everyone uz , z 1, Z are chosing matrix is formed
hz* {max hiNz ,i 1, n}
Cnd * [hz* ] [(h1*, h1*, h1*, h1* ), ... , (hZ* , hZ* , d Z* , hZ* )] According to the words (11) hz* 1 for z , that is, all elements of the matrix Cnd * are equal to units.
Step 8. The ideal negative (worst) solution is calculated (INS) Cnd . To this aim, for everyone uz , z 1, Z are chosing and the following matrix is formed:
ez {min eiNz , i 1, n}
Cnd [ez ] [(e1 , e1 , e1 , e1 ), ... , (eZ , eZ , eZ , eZ )] Step 9. Using the formula (3), according to the individual values of each individual criterion, the distance of the applicants to IPS: On the basis of the obtained results a vector is formed [H * ] [H1* ,..., H Z* ] . Step 10. By the individual values of each individual criterion, the distance of the applicants to INS: (11) (12) (13) (14) (15) H z (cnd i , Cnd )
((eiNz ez )2 ( fizN ez )2 (giNz ez )2 (hiNz ez )2 ) (17) On the basis of the obtained results a vector is formed [H ] [H1 ,..., H Z ] . Step 11. Determine the distance of each of the applicants to IPS and INS: H * (cndi )
Z (H z* (cnd i ,Cnd * )2 ; D (cnd i ) z1
Z (H z (cndi ,Cnd * )2 . z1 (18) Step 12. Calculates the integral index (coefficient of proximity) for each comparable bidder as the ratio of the calculated distance for him from the ideal worst decision to the sum of distances to the best and worst decisions:
H (cnd i ) H * (cnd i ) H (cnd i ); (cnd i )
H (cnd i ) .
H (cnd i ) (19) In accordance with the value of the coefficient of proximity (cnd i ) it is possible to rank applicants. So, the closer to the unit value of the coefficient of proximity (cnd i ) , the better the candidate is compared. The largest value of the integral indicator (cvdi ) determines the best bidder, that is, the optimal solution. The least value (cnd i ) matches the worst bidder. 5
Conclusions In the conditions of the active development of innovative information technologies, human resources, including IT specialists, have become the main strategic resource of companies, which ensures their long-term competitiveness and achieve their goals. The emphasis in this matter is on recruiting IT specialists. In addition to the basic requirements for professionalism and IT education, the latest trend in recruiting is to take into account the psychological portrait of a future employee as a level of communicativeness, openness of experience, good faith, goodwill, leadership qualities, etc. One of the modern tasks of analyzing the psychological state of personality when recruiting IT specialists is to analyze his profile in social networks (posts on forums, comments on events, activity in social networks, etc.). Therefore, the development of new conceptual approaches and promising IT management of human resources becomes of special relevance and practical significance. In recruiting, the challenge of recruitment is of great importance, since, with the presence of qualified personnel, the company can function successfully. Therefore, the issues of the adoption of personnel decisions, free of subjectivity, are very relevant. The complexities faced by companies in the process of identifying the applicant, the most acceptable requirements for a particular post in terms of professional suitability, and from the point of view of compliance with corporate style and psychological compliance, necessitate the development and improvement of automated approaches for recruiting. The paper proposes a methodological approach to solving recruiting tasks based on the analysis of data on the applicant from social networks using multi-criteria optimization based on the TOPSIS method. The use of the TOPSIS method in human resources management tasks allows to increase the adequacy of the decisions made by ranking according to the degree of proximity to the ideal solution ensures the objectivity and transparency of the adopted managerial decisions and provides opportunities for expanding the scope of multicriteria optimization methods. The authors apply the TOPSIS method modification, which consists in integrating into the algorithm an additional step, which involves calculating and introducing the reliability coefficients of social networks involved in the procedure for evaluating the applicants' psychological state based on their activity with emotional coloration (neutral, positive and negative).
The advantages of the proposed approach to multicriteria optimization based on the modified TOPSIS method to support decision-making in recruitment are as follows: Absence of the need to compile a fuzzy rule base; Mathematical reasonableness and relative simplicity of calculations of integral indicators, allowing to carry out ranking of alternative solutions, to carry out further analysis and choice of the final solution; Lack of constraints on the number of applicants and the parameters of the formation of criteria for evaluation according to the activity in social networks, which characterize the object of research; Accounting in the algorithm of making decisions on the competence of the years of training involved in the decision-making process; Accounting for the hierarchical structure of the criteria describing the applicants; Ability to rank applicants according to their degree of proximity to the ideal solution.
The article presents a step-by-step demonstration of the possibilities of the proposed method in the process of multi-criteria analysis and decision-making on the example of the selection and recruitment task. Conducting alternative calculations for decisionmaking based on a ball assessment system and a comparative analysis of the results of the two methods shows the effectiveness of the proposed method. Using the described methodological approach as the mathematical basis of a computer system for decision support in recruiting tasks can become an effective tool for preparing and taking effective decisions in human resource management in IT business. 69. Vitynskyi, P., Tkachenko, R., Izonin, I., Kutucu, H.: Hybridization of the SGTM Neurallike Structure through Inputs Polynomial Extension. In: International Conference on Data Stream Mining & Processing (DSMP), 386-391. (2018) 70. Tkachenko, R., Izonin, I., Kryvinska, N., Chopyak, V., Lotoshynska, N., Danylyuk, D.: Piecewise-linear Approach for Medical Insurance Costs Prediction using SGTM NeuralLike Structure. In: Informatics & Data-Driven Medicine (IDDM), 170-179. (2018) 71. Tepla, T.L., Izonin, I.V., Duriagina, Z.А., Tkachenko, R.О., Trostianchyn, А.М., Lemishka, І.А., Kulyk, V.V., Kovbasyuk, Т.М.: Alloys selection based on the supervised learning technique for design of biocompatible medical materials. In: Archives of Materials Science and Engineering, vol. 93/1, 32-40. (2018) 72. Izonin, I., Trostianchyn, A., Duriagina, Z., Tkachenko, R., Tepla, T., Lotoshynska, N.: The Combined Use of the Wiener Polynomial and SVM for Material Classification Task in Medical Implants Production. In: International Journal of Intelligent Systems and Applications (IJISA), vol. 10(9), 40-47. (2018) 73. Tkachenko, R., Duriagina, Z., Lemishka, I., Izonin, I., Trostianchyn, A.: Development of machine learning method of titanium alloys properties identification in additive technologies. In: EasternEuropean Journal of Enterprise Technologies, vol. 3, 12(93), 2331. (2018) 74. Mahmud, J.: IBM Watson Personality Insights: The science behind the service. In: IBM Corporation, https://developer. ibm. com/watson/blog/2015/03/23/ibm-watson-personalityinsights-science-behind-service. (2015) 75. Mammadova, M. H., Jabrayilova Z. H.: Multi criteria optimization of human resource management problems based on the modified topsis method. In: Eastern-European Journal of Enterprise Technologies, vol. 2(4 (74)), 48-62. (2015) 76. Mammadova, M. H., Djabrailova, Z. G.: Decision-Making Support in Human Resource Management Based on Multi-Objective Optimization. In: TWMS Journal of Pure and Applied Mathematics, vol. 9(1), 52-72. (2018) 77. Calvo-Armengol, A., Jackson, M. O.: The effects of social networks on employment and inequality. In: American economic review, vol. 94(3), 426-454. (2004) 78. Chen, C. T., Hung, W. Z.: Choosing project leader based on interval linguistic TOPSIS and social network technology. In: 2012 International conference on Fuzzy Theory and Its Applications (iFUZZY2012), 310-315. (2012) 79. Clais, P., Leseur, G., Lasfar, O.: U.S. Patent Application No. 14/527,548. (2015) 80. Kotsilieris, T., Pavlaki, A., Christopoulou, S., Anagnostopoulos, I.: The impact of social networks on health care. In: Social Network Analysis and Mining, vol. 7(1), 18. (2017).