=Paper=
{{Paper
|id=Vol-2258/paper14
|storemode=property
|title=Financial Sustainability Evaluation of Higher Education Institutions Using Compatible Cognitive Maps
|pdfUrl=https://ceur-ws.org/Vol-2258/paper14.pdf
|volume=Vol-2258
|authors=Yaroslav Fedulov,Vadim Borisov,Alexander Fedulov
}}
==Financial Sustainability Evaluation of Higher Education Institutions Using Compatible Cognitive Maps==
https://ceur-ws.org/Vol-2258/paper14.pdf
Financial sustainability evaluation of higher education
institutions using “compatible” cognitive maps
Y A Fedulov1, V V Borisov1 and A S Fedulov1
1
Computer Engineering Department, The Branch of National Research University “Moscow
Power Engineering Institute” in Smolensk, Smolensk, Russia
Abstract. The paper considers the features of monitoring the effectiveness of financial and
economic activities of higher education institutions. Integral (group) indicators are identified
and the main approaches to obtaining evaluations for them are considered. The distinctive
features of the financial sustainability group indicator are considered. The evaluation method
of the financial sustainability indicator for higher education institutions is presented on the
basis of fuzzy cognitive modeling taking into account the different compatibility levels
between the concepts and different weights of mutual influence. A set of convolution
operations corresponding to various degrees of compatibility between concepts has been
selected and justified. Based on the results of the simulation, it is possible to determine: the
facts of exceeding by a certain moment the concepts criterial values; predictive values and
trends in the values of different concepts; effects evaluation of direct and mediated external
influences on concepts.
1. Introduction
The effectiveness of financial and economic activities of higher education institutions is determined
largely by the quality of the planning organization and budgeting processes in the institution. In order
to control the effectiveness of spending budget funds, the Ministry of Education monitors a number of
integrated (group) indicators of financial and economic activity, including:
panning quality indicator;
financial sustainability indicator;
strategic planning indicator;
distribution of the organization's funds;
accrued salary fund;
scientific research costs.
As a methodological support for obtaining evaluations for the selected integrated indicators, it is
proposed to use two main approaches.
Application of deterministic algorithms that provide an exact solution. These algorithms allow
to obtain values of integral quality indicators based on averaging of weighted partial
evaluations (average values of monitoring indicators or average value of the final quality
indicator depending on the task setting). This approach to obtaining estimates of the financial
and economic activities effectiveness is used for most group indicators.
Application of fuzzy cognitive modeling algorithms to obtain values. The fuzziness can be
caused by the absence in some cases of input primary data exact values (monitoring
indicators), for example, due to the lack of indicator values for the estimated period (quarter,
102
month) in the information array. The predicted values are characterized by uncertainty,
probability; subjective assessments of experts can be applied. Described types of uncertainty
can be found in the financial sustainability integral factor, which is one of the main monitoring
indicators of the higher education institutions financial activities effectiveness.
Actively developed methods of cognitive analysis [1-3] and modeling provide advanced
opportunities for qualitative and quantitative analysis and modeling under conditions of stochastic and
non-stochastic uncertainty, allowing to solve a wide range of both analytical problems [4] (stability
analysis, identification of undesirable cycles, evaluation of system characteristics, analysis of direct ,
aggregated and mediated influence of system factors on each other, scenario analysis under various
impacts, and the availability of target situations, the forecast of the development of the state of
systemic factors), and modeling tasks [7,8] (modeling of system dynamics without and in the presence
of external influences, modeling states in conditions of limited resources).
At the same time, in the case of the of financial sustainability indicator, it is also necessary to take
into account the compatibility of system factors - concepts - both in determining the direct influence
between the concepts, and in accumulating the influence of several concept-sources on the concept-
receiver. But, simultaneously, the accounting mechanism of the systemic factors compatibility,
realized in generalized production fuzzy cognitive maps: first, it is oriented only to the representation
of the concepts meanings and relations between them in the form of linguistic terms; secondly, it is
quite complicated for formalization and expert interpretation; thirdly, it requires rather complicated
procedures for constructing production rules for modelling system dynamics.
The paper considers the method that allows to generalize the solution of the problem of the system
factors compatibility accounting with different levels of influence between them on the example of the
higher educational institutions financial sustainability indicator.
2. Fuzzy cognitive model for financial sustainability integral indicator
A fuzzy cognitive map, taking into account the compatibility of its concepts, represented as follows:
FCM K , W , C ,
where K {K1 , ..., K N } – set of concepts; W {wij } – set of concepts influence weights on each
other; C {cij } – set of compatibility levels of concepts pairs, i, j = 1, …, N [9].
Depending on the nature of problem being solved, [5] compatibility of concepts can be interpreted
as correlation, mutual influence, simultaneous achievement of their criterial values, etc.
Hereinafter, the compatibility levels of map concepts are taken into account in the choice of
operations both for determining the direct spread of influence between concepts, and for accumulating
the influence of several concept-sources on the concept-receiver.
As an example of using the proposed method of fuzzy cognitive modeling to obtain evaluations, the
construction of a possible compatible cognitive model for calculating the integral financial
sustainability indicator based on the accumulation of the influence of private concepts is considered.
Based on the normative documents regulating private indicators of monitoring the educational
institutions financial activities, in constructing a fuzzy cognitive model, a structure of concepts was
formed, consisting of financial sustainability integral indicator of educational institution K 6 (target
concept); and five private concepts of a compatible cognitive map that influence one another and
directly on the target concept:
K 1 – the share of income from income-generating activities in the total income from income-
generating activities and subsidies for financial support for the fulfillment of the state task
(autonomy indicator);
K 2 – the increase in income from income-generating activities in relation to the previous
year;
K 3 – institution dependence on borrowed financing sources (debt ratio);
K 4 – the presence of the overdue accounts payable;
103
K 5 – the presence of accounts receivable to collection.
The structure of the obtained fuzzy cognitive model of the financial sustainability integral indicator
is presented on figure 1.
K1 w
51 ,c
, c 13 , c 21 w 51
w 13 w 21 c 12 15 ,c
15
,
31
w 12
,c
w52 , c52
1
w3
w14, 1
w61, c61
K5
w16 , c16
K2 w62 , c
c4
62
w , c 56
26 ,c w 56
26
K6
w32 , c32
53
w 24
,c
w23, c 23
,c 2
w 24
53
w
w46 , c 46
4
, c 24
45
,c
45
w
,c 3
6
w 36
35
w34 , c34
,c
K4
35
w
K3 w43 , c43
Figure 1. Fuzzy cognitive model of the financial sustainability integral indicator.
3. Construction of the mutual influence weights table
Concepts meanings and their mutual influences representation in the form of fuzzy numbers, sets, and
relations makes it possible to use for their evaluation methods of fuzzy arithmetic, fuzzy sets and
relations, fuzzy logic for their evaluation.
Consider, for example, fuzzy cognitive maps, characterized by the following features [5]. Concepts
K i take values from the range [0, 1]. The levels of influence between each concept pair K i and K j
from K {K1 ,..., K N } are represented in the form of directed arcs with weights wij [1, 1] .
Value wij 1 characterizes the most negative, and wij 1 – maximally positive impact of the
concept K i on concept K j , and value wij 0 means the absence of direct influence of the concept
K i on concept K j . With the positive influence of the concept K i on the concept K j ( wij 0 ) and
increasing K i means that the value K j grows, and decreasing – diminish. When wij 0 value
increasing of K i will reduce the value K j and, vice versa.
It is also possible to simultaneously influence the pair of concepts against each other with different
values wij and w ji .
According to the expert survey results, the mutual influence weights of the considered concepts are
represented on table 1.
104
Table 1. Mutual influence weights of private indicators.
№ Indicator 1 2 3 4 5 6
1 K1 0 0.85 0.72 -0.22 -0.34 0.87
2 K2 0.63 0 0.44 -0.31 0 0.69
3 K3 -0.19 0.40 0 -0.12 -0.21 -0.35
4 K4 0 -0.43 -0.36 0 -0.36 0.56
5 K5 0.45 0.44 0.47 0 0 0.58
6 K6 0.91 0.79 0 0 0 0
4. Construction of the mutual influence weights table
To determine the compatibility levels of the fuzzy cognitive map concepts, both direct and inverse
methods can be used. When using direct methods, these values are either set by experts or obtained as
a result of experiments.
Inverse methods are used in the case of the direct evaluation complexity of the concepts
compatibility levels. For example, a method based on a pairwise comparison of directly interacting
concepts and filling the so-called compatibility tables between the linguistic values of these indicators
can be applied [10].
For fuzzy cognitive maps considered in [2], when assessing the concepts compatibility levels, one
can also use the results of the relationships analysis of their mutual influence on the basis of indicator
of consonance (coherence) of the concepts influence on each other.
Based on the compatibility tables analysis for all pairs of private concepts, six criteria compatibility
levels are defined: {1 – “No compatibility”, 2 – “Low level”, 3 – “Below average level”, 4 – “Above
average level”, 5 – “High level” 6 – “Full compatibility”}.
Table 2 presents a fuzzy relation of the concepts compatibility of the cognitive model in question as
a compatibility matrix.
Table 2. Fuzzy compatibility matrix.
№ Indicator 1 2 3 4 5 6
1 K1 – 6 5 2 3 6
2 K2 5 – 4 3 4 5
3 K3 2 3 – 1 1 3
4 K4 – 3 3 – 3 4
5 K5 4 4 4 – – 4
6 K6 6 5 – – – –
5. The fuzzy convolution operations basis
To directly spread influence from the concept K i to the concept K j as operations of fuzzy influences
“weighing” (considering the compatibility of these concepts) it is advisable to use the family of
parametrized operations med, which satisfy the axioms of normalization, nondecreasing, continuity,
bisymmetry [6]. Depending on the problem situations nature, the following expressions can be used:
K j med( Ki , wij ; cij ); K j K j med( Ki , wij ; cij ); K j K j med( Ki , wij ; cij ).
where K j – state (value) of the j-th concept-receiver; K i , K i – state (value) and increment of the i-
th concept-receiver; i, j {1, ..., N }, cij [0, 1] .
To compare the compatibility levels of aggregated concepts with the operations of convolution, as
a rule, direct methods are used, in which the expert establish such conformity.
105
When evaluating and selecting alternatives, different strategies can be used, the extreme variants of
which are: achieving the lowest value from all indicators or reaching a maximum value for at least one
of the indicators.
For the two-place case, the following convolution operations correspond to these extremal
strategies min(k , k ) med(k , k ; 0) and max(k , k ) med(k , k ;1.0) .
Thus, it can be concluded that the whole set of compromise strategies provides a parametrized
family of convolution operations of the type: med(k , k ; ), [0, 1].
To determine the degree of compromise of two-fold convolution operations, it is proposed to use
the parameter [0,1] , in this case, the smaller the value of the parameter, the less the degree of
compromise of indicators.
The values of the parameter are calculated in accordance with the expression:
v vmin
,
vmax vmin
where v – the “volume” value under the surface of the function formed as a result of applying the
corresponding convolution operation; and а vmin , vmax – as a result of operations
min(k , k ) med(k , k ; 0) and max(k , k ) med(k , k ;1.0) , respectively.
Parameter values can be used to determine parameter of this convolution and characterize the
compatibility levels of the aggregated concepts. Figure 2 shows the dependence on the .
1
0,95
0,84
0,65
0,5
0,35
0,16
0,05
0
0,1 0,2 0,25 0,3 0,4 0,5 0,6 0,7 0,75 0,8 0,9 1
Figure 2. The dependence of on the for the med(k , k ; ) convolution operation.
Using the proposed approach, the selected compatibility levels of indicators were compared to the
following convolution operations:
h1 (k , k ) med( k , k ; 0) h3 (k , k ) med( k , k ; 0.43) h5 (k , k ) med( k , k ; 0.71)
h2 (k , k ) med( k , k ; 0.29) h4 (k , k ) med( k , k ; 0.56) h6 ( k , k ) med( k , k ; 1.0)
The set of operations directly influencing the concept K i to the concept K j considering the
compatibility levels of these concepts is represented on the table 3.
106
Table 3. The set of operations directly influencing the concept K i to the concept K j
considering the compatibility levels of these concepts.
№ Direct influence propagation Criterial compatibility levels
operation from K i to K j Designation Description
1 K j med( Ki , wij ; 0.0) NC No compatibility
2 K j med( Ki , wij ; 0.29) LC Low level
3 K j med( Ki , wij ; 0.43) BC Below average level
4 K j med( Ki , wij ; 0.56) AC Above average level
5 K j med( Ki , wij ; 0.71) HC High level
6 K j med( Ki , wij ; 1.0) FC Full compatibility
6. The strategies for the influences accumulation
If the number of private concept-sources is more than 2, the result may depend on the order of
accumulation of their influences.
The following strategies for accumulating the effects of several concept-sources on one concept-
receiver are applicable: either from the least to the greatest compatibility level of concepts, or from the
greatest to the least compatibility level of concepts.
Thus, for the example presented in Figure 1, the expression for accumulating the concepts
influence on the target concept of the financial sustainability integral indicator (at the accumulation
1 2 3 4
order from the least to the greatest concepts compatibility level (((( K3 K4 ) K1 ) ) K5 ) K2
takes the form:
K6 K 2 + med K5 med K1 med(( K3 med( K 4 , w34 ; 0)), w13 ; 0.29) , w51;0.43 , w25 ;0.56 .
The concepts compatibility here is interpreted as the simultaneous attainability of the criterial
concepts values.
Modelling the influence propagation on the map can be performed under conditions of self-
situation, or with external force moments at various values of concepts, which all reduces to the
following sequence:
setting the initial concepts values fuzzy cognitive map;
the simulation start in accordance with the selected expression to change the concepts values;
setting the influences on the map concepts at certain moments (changing their values);
completion of the simulation when the selected criterion is fulfilled.
Based on the results of the simulation, it is possible to determine: the facts of exceeding by a
certain moment the concepts criterial values; predictive values and trends in the values of different
concepts; effects evaluation of direct and mediated external influences on concepts.
7. Conclusion
The article considers the features of monitoring the effectiveness of financial and economic activities
of higher education institutions. Integral (group) indicators are identified and the main approaches to
obtaining evaluations for them are considered. The distinctive features of the financial sustainability
group indicator are considered.
An approach allowing to generalize the problem solution of the system factors compatibility
accounting in the transfer of influence between them for different fuzzy cognitive maps is considered
107
on the example of the financial sustainability integral factor. A set of convolution operations
corresponding to various degrees of compatibility between concepts has been selected and justified.
8. References
[1] Kosko B Fuzzy cognitive maps Int. Journal of Man-Machine Studies, vol. 24, 1986, pp 65-75.
[2] Silov V B Strategic decision making in fuzzy environment Moscow, INPRO-RES, 1995. (in
Russian)
[3] Carvalho J P, Wise L, Murta A, Mesquita M Issues on Dynamic Cognitive Map Modelling of
Purse-seine Fishing Skippers Behavior IEEE World Congress on Computational Intelligence,
WCCI2008-2008, Hong-Kong, 2008, pp 1503-1510.
[4] Zhang, Shanwen, Harry Wang, Wenzhun Huang Two-stage plant species recognition by local
mean clustering and Weighted sparse representation classification. Cluster Computing
(2017): pp 1-9.
[5] Borisov V V, Fedulov A S Generalized Rule-Based Fuzzy Cognitive Maps: Structure And
Dynamics Model Lecture Notes in Computer Science, 2004, v. 3316, pp 918-922.
[6] Papageorgiou E I, Hatwagner M F, Buruzs A, Koczy L T A concept reduction approach for
fuzzy cognitive map models in decision making and management Neurocomputing, vol. 232,
pp 16-33.
[7] Borisov V V, Dli M I and Kozlov P Yu 2017 Intellectual Methods of Nonstructural Texts
(Smolensk: Universum). – p 156 ISBN 978-5-91412-364-9 (in Russian)
[8] Guichao F, Denghua Z, Fugen Y, Pan Y A hybrid fuzzy evaluation method for curtain grouting
efficiency assessment based on an AHP method extended by D numbers Expert Systems with
Applications 44 (2016): pp 289-303.
[9] Borisov V V, Fedulov A S, Fedulov Y A “Compatible” Fuzzy Cognitive Maps For Direct And
Inverse Inference Proc. of the 18th Int. Conf. on Computer Systems and Technologies,
CompSysTech’17, June 23-24, Ruse, Bulgaria, ACM Int. Conf. Proc. Series, Vol. 1369, ACM
Inc., N.Y. USA, pp 20-27.
[10] Zernov M M Method of designing of a fuzzy multi-criteria evaluation models Nejrokompyutery:
razrabotka, primenenie Neurocomputers: development, application, 2007, no 1, pp 40-49. (in
Russian)
Acknowledgments
The reported study was funded by RFBR according to the research project № 18-29-03088.
108