QUANTITATIVE MODELING OF THE RECRUITMENT PROCESS

• Main Author: A. Zinchenko
• Format: Article
• Language: Russian
• Published: Government of the Russian Federation, Financial University 2015-09-01
• Series: Управленческие науки
• Subjects: recruitment; binary regression; artificial neural network; personnel management
• Online Access: https://managementscience.fa.ru/jour/article/view/34

The article presents a quantitative approach to modeling the process of recruitment. In circumstances where it is necessary to collect a large staff in a short time, the use of mathematical apparatus can significantly accelerate this process. The most complex and labor-intensive, but also ambiguous on the result, is a step in which the direct decision on hiring is made. Interviewing and resume viewing are always associated with human subjectivity. In addition, it is not so easy to find the recruiter with sufficient professional experience, hold him, motivate, etc.Modern works dedicated to recruiting, describe the process of making a decision on hiring from the point of view of a manager, who assesses professional and personal qualities of the future employee. The author proposed a method for estimating the probability of passing the probation, based on the information provided in the CV. To solve this problem, author used a binary choice model and artificial neural network and made a comparative analysis of these models. Advantages of each approach may occur depending on the situation. For example, coefficients in binary regression let you know exactly how certain factors affect the probability of passing the probation period. Such information can be usefulin a situation where the decision on hiring employees in managerial positions is made and deep analysis is needed. On the other hand, binary regression has some drawbacks: in some cases the binary choice model can produce uncertain evaluation of candidates.This problem does not occur when using the ANN to the task of evaluating candidates for the position: ANN provides an unambiguous classification. However, synaptic weights in the artificial neural network cannot be interpreted, making it impossible to determine the effect of individual factors on the results provided by the model.