This article presents an approach to the feasibility of investing in human capital in the framework of the development of socio-economic systems of intellectual capital. In conditions of partial or complete uncertainty in the presence of incomplete input data, the use of tools based on artificial intelligence methods (neural networks, genetic algorithms) allows us to evaluate a wide range of indicators. The article presents non-linear mathematical models for assessing the feasibility of investing in the intellectual capital of a university, in particular, in its structural component - individual human capital, namely, in the formation of an individual. Models are based on logistic regression and neural networks. The authors proposed several key factors affecting the binary resulting variable "investment effect": the level of knowledge and skills in the specialty; self-education (in terms of regularity of raising the level of information culture); employee age. The developed methodology is the basis of a decision support system that allows for the strategic development of the university based on effective investment in the human capital of its employee.
In the conditions of the digital transformation of all spheres of human life, his creative and intellectual potential acquires a new significance. The future of humanity never then earlier depends on competent balanced decisions, for every person's potential, which is unique, must be more fully disclosed and optimally used. But the reality is that the streaming development of the digital economy is ahead of the willingness of each individual to comply with the newest technologies - it needs permanent growth, continuing education.
Human capital development may be a more important factor in the long-term success
of a country (region or any large-scale system) success. Ingenuity, creativity, and an
imaginative approach, which has humanity, allow us not only to solve the main
problems of our time but to build a future, which is oriented on a human being [
According to the report about human capital in 2017, which is developed by the
World economic forum, a global index of human capital is estimated for 130 countries
by the scale from 0 to 100 by four sub-indexes (potential, development, deployment,
and know-how). It aims to provide a holistic assessment of the human capital of the
country, both current and expected values. It allows doing an effective comparison
according to regions, age groups, the income of people, etc. [
According to the national educational doctrine in the Russian Federation up to 2025
year, one of the main aims is “lifelong persons steadily education” [
The problems of developing human capital and evaluating the effectiveness of
investments in human capital, in particular in education, were studied by such scientists: G.S.
Becker [
The intellectual capital (IC) market cannot satisfy the investor according to somewhat algorithm. Except for unique genius, many talent specialists or simply obligatory specialists in a certain area, all other potential employees only with certain probability can impact on the organization profitability, by investing capital in them, or else, by investing in intellectual recourse.
However, the feasibility of investing in individual human capital is relevant.
Even a return from geniuses and talents depends on the vagaries of the "human factor" (for example a person loses his motivation at a certain stage of his intellectual potential growth, or transmission of values exists). In the human capital market is observed a constant rivalry: the income of each investor often depends on how many other employers are willing to invest in a certain moment of an uncertain future in human capital. But no one can control or even predict the behavior of a large number of other investors with a sufficient degree of reliability. However, the employers, who are investing in the human recourse, or generally, in the intellectual capital, can carry on a risk, which they take on themselves.
The research will focus on the evaluation of an investment in teachers of higher education organizations, when educational institutions seek to build capacity for strategic development, given the dynamics changes especially.
During the formation of the “investment portfolio” IC of the organization appears a problem of evaluation of return probability from the investment or the task of evaluating risk. In the frames of this task conclusion, it might use logistic binary regression.
Let F1, F2, …, Fn – be a factor sign, Y – resulting binary sign: Y 0;1. We set ourselves the task of determining the object of study belonging to one of two classes.
Let give examples of binary regression using: ─ issuance of credit – refusal to issue a credit – the solving of a credit scoring task based on the values of the factors F1, F2, …, Fn, specified in the application. The resulting sign value, which takes only two values: Y=1, if the client is trustworthy, and Y=0 otherwise, is predicting; ─ link operators – the solving of problems like payable account: customer payment, wherein Y=1, if the customer is not going to change the operator, and Y=0 – client transition to contestants; ─ equipment replacement task: according to the instrument readings F1, F2, …, Fn, the task of predicting an uninterrupted operation of equipment for a certain time is solving; ─ advertising response: wherein Y=1, if the customer has bought some product or service, and Y=0 – the potential client hasn’t bought anything.
In the case of solving an investment in the intellectual capital task, particularly in human capital, it necessary to check concrete factor signs, which characterize an investment object. It must be noted, they are chosen purely for a particular organization, taking into account its strategic objectives and tactical measures. Among that factors maybe sex, intellect level, creativity level, participation in conferences, marital status, number of children, housing conditions, etc.
We can recognize several main factors that influence the parameter being evaluated: ─ the level of knowledge and skills in the specialty (F1); ─ self-education (in terms of regularity of information culture level increasing - F2); ─ age of employee (F3).
There are presently only three factors among the main ones in this research, affecting the impact of return on investment in the staff of the department. There may be much more factors, their quantity is determined by the researcher. The meaning of the selected factor F1 is shown in Fig. 1.
The meaning of the selected factor F2 is shown in Fig. 2.
It can be noted, that the evaluation of the level of knowledge and skills in the specialty and self-education can be carried out by an expert method (experts are department staff himself, his colleagues, the head of the department). (1) (2) P(Y | F )
1 e(b0 b1F ) P(Y | F ) 1 e(b0 b1F1 b2F2 b3F3 ) 1 1 As a binary resulting variable in this task can be chosen an effectiveness of department staff activity (“effect” – “return”). This variable will take the meaning Y=1, in case it appears effect from investments (copyright certificates, patents, publications in Scopus, Web of Science, etc.), and Y=0 if the investment effect is absent.
For evaluation of the impact of mentioned above signs on the resulting binary, the
variable can be used the mathematical modeling method, based on the using of the
logistic regression equation. The logistic model allows evaluating a probability
P(Y | F ) of dependent variable (Y) binary result independence from the independent
variable (F) meanings according to the next formula [
In the case of existing of three independent variables it may use the next formula: The coefficients of the model are calculated by using instruments of the STATISTICA package (the module Nonlinear Estimation can be used for these purposes), presented in Fig. 3.
The calculation of the model parameters is presented in Figure 4, and evaluation of the model parameters – is in Figure 5.
Accordingly, for the initial data, one can use a formula with the following logistic regression coefficients, the calculation results of which are presented in Table 1.
P(Y | F )
1 1 e (35,3750,793F13,700F2 0,334F3 ) (3) A conclusion is made by the calculated probability value P(Y | F ) 0,5 , that the predicted value of a dependent variable must to be equal Y=1. The otherwise conclusion (Y=0) is made in the case of the calculated probability value P(Y | F ) 0,5 .
Based on data, outlined in Table 1 (a factual value of dependent variable Y and its predicted values), it can be resumed, the accuracy of the developed logistic model is about 95%, it has correctly predicted 19 values of the dependent amount.
Recently, artificial intelligence methods have become widely used [
Among the obvious advantages of using neural networks, it can be highlighted a large sector of the use of this tool in various branches of science and production (data analysis, optimization, forecasting, marketing, and advertising, work with images, audio, and video resources, etc.).
Also, it must be highlighted the ability to filter input data, fast learning of neural networks, detection of symptoms when approaching a critical point. The calculated neural networks allow fulfilling the modeling according to incoming data accurate to 100% (as shown in Fig. 7). Let us consider the network, which is selected on the minimum error criterion. It has a formula MLP 3-3-1. This network represents a multilayer perceptron, consisting of 3 neurons in incoming and hidden layers and one neuron in the outlet layer. The calculated values of weight coefficients for this network are presented in Fig. 8.
As shown in Fig. 9, the values of input signs, output binary variable, and calculated values of network exit suggest, that the developed neural network has correctly predicted all 20 values of the dependent variable. Also, it should be noted, that software package STATISTICA tools make it possible to calculate values of output binary dependent variable for various user’s values of input variables, not only for those present in input data.
According to the comparison of the results, received by logistic regression, and that, received by the neural network, a conclusion is made, that using neural networks is a more accurate modeling method of complete nonlinear dependencies.
An approach with using neural networks may be difficult to a certain degree for those specialists, ho not able to use these tools. The use of a logistic regression equation is one of the more simple modeling methods of binary dependencies. Its characteristics can be obtained even as a result of calculations, for an instant, in an Excel spreadsheet.
A set of factors, in addition to those chosen by the authors, may also contain such factors, as the level of culture of scientific activity, the level of methodical culture, the level of educational activity, sex, the presence of children in employees (their number), the gradation of children by age categories (for example, up to 3 years, from 3 to 7, primary school age, etc.), the availability of own housing, living conditions, etc.
The developed method is the basis of the decision support system that allows strategic development of high education organizations based on effective investment in the human capital of its employees.