Information Support of Intelligent Decision Support Systems for Managing Complex Organizational and Technical Objects Based on Markov Chains

Abstract

Management of multilevel organizational and technical systems under the influence of environmental factors is a complex process that uses both structured and semi-structured data. For information support of management decisions in such systems, the use of probabilistic mathematical models based on Markov processes is proposed. In contrast to the traditional use of Markov chains, it is proposed to replace equal step intervals with a discrete sequence of states determined by environmental influences. This approach makes it possible to model and regulate the process of making relevant decisions when managing multi-level organizational and technical objects and increase its efficiency in difficult operating conditions. information support, intelligent systems, semi-structured problems, control, uncertainty, COLINS-2022: 6th International Conference on Computational Linguistics and Intelligent Systems, May 12-13, 2022, Gliwice, Poland. ORCID: 0000-0003-2321-459x (M. Sharko); 0000-0001-7383-8558 (N. Petrushenko); 0000-0003-3917-7586 (O. Gonchar); 0000-0001-79105013 (N. Vasylenko); 0000-0002-3990-730X (K. Vorobyova); 0000-0003-0918-9949 (I. Zakryzhevska)

1. Introduction

The typology of solving complex semi-structured problems of managing multilevel control systems requires taking into account quantitative and qualitative characteristics with the dominance of uncertainty and fuzzy ideas about the influence of unpredictable environmental factors. The appearance of a hierarchical structure in intelligent semi-structured control systems is due to the presence of a large amount of information about the controlled processes in the system, the impossibility of processing this information and making decisions by one control center, as well as the decentralization of the decisionmaking process. One of the important problems of decision-making under conditions of uncertainty is the lack of a common methodology for constructing probabilistic models for regulating the process of making relevant decisions and information support for intelligent control systems for complex multilevel organizational and technical objects.

Decision-making information support models in the management of simple organizational and EMAIL: (M.

Sharko); [e-mail] (O.

Gonchar);

2022 Copyright for this paper by its authors. technical objects are used to analyze systems in which decision-making is of a one-time nature, and system components are described by static quantities. Most management models for complex organizational and technical objects are characterized by the fact that the processes they describe are dynamic in nature. Dynamic models of hierarchical control systems for complex organizational and technical objects operating under conditions of uncertainty are of particular interest due to the need to take into account controllable and uncontrollable factors.

Phenomenologically, the choice of the first step to change the current situation of managing complex organizational and technical systems is associated with a quantitative assessment and adjustment of one of the determining factors, which leads to a shift in the starting point of the management transformation process. After performing operations related to the adjustment of the subsequent factor, the starting point will again shift towards the reduction of the process. Thus, the process of changing the position of the reference point is random in nature, characterized by an arbitrary choice of a corrected factor with discrete time characteristics of the duration of the first and subsequent steps and a countable set of states. Such a process will be Markovian, since subsequent states of the starting point of the process of transformational transformations do not depend on past states.

The unresolved parts of the general problem of managing complex objects include the formalization of the accumulated knowledge and experience in managing them, taking into account the influence of uncertain destabilizing environmental factors on the cognitive component of the decision maker.

The aim of the work is to develop information support tools for intelligent decision-making systems in the management of multi-level organizational and technical objects under conditions of uncertainty.

2. Relative Works

The Markov process model applied in the construction of logical networks of the information space is presented in [1,2]. In [3], the Markov process model is used to calculate the probabilities of transitions between states of patients as a set with deviations in the anatomy of the lymphatic drainage system. The construction of the architecture of information support for the management of organizational and technical systems is presented in [4]. The use of the theory of artificial intelligence and computational linguistics to enhance the semantic connection of uncontrolled terms of knowledge representation in information systems of engineering regulations is presented in [5,6]. The procedure for information support of decision-making systems with a limited number of observations, based on interval estimates of probabilities, is presented in [7]. The information-entropy model of the basis for making managerial decisions under conditions of uncertainty is presented in [8]. The information support system for minimizing losses in the management of information systems with the analysis of the results of management decisions is presented in [9]. In [10], a description of information support for managing uncertainty by taking into account the requirements, opportunities and recommendations for the implementation of projects in corporate systems is given. Information support of mechanisms and types of control with a gradation of categories of possibilities and classes of uncertainties is presented in [11]. The use of information technologies for risk and uncertainty management in complex projects is presented in [12,13]. The influence of information systems on business efficiency is reflected in the works [14-17]. In [18], a number of new functions and influences on management activities are proposed. The adoption of preventive measures to minimize threats and risks is reflected in [19]. The work [20] is devoted to setting priorities and reducing uncertainties by digital data transformation. The use of a heterogeneous hidden Markov chain for the characteristics of wavelet coefficients is considered in [21], feature extraction based on the Markov chain for anomaly detection in time series in [22], the use of Markov chains in complex multilevel control chains in [23]. Modeling using Markov chains for a wide range of applications is presented in [24-28].

The variety of ways to study uncertainty has caused fragmentary ideas about the parameters of uncertainty and approaches to its management, inconsistency in conceptualization and measurements. The formation of a modern information support system in system research should be aimed at modernizing the tools for managing organizational and technical systems in the context of dynamic transformations.

3. Materials and Methods

The properties of information support for decision-making under conditions of uncertainty were used as research materials:  Information security  Protection from the influence of the external environment  Controllability, i.e. the possibility of adjusting control actions  Structural heterogeneity, i.e. the presence in the system of various elements with different weight contributions  Effectiveness, i.e. the emergence of a new quality in the combination of a specific set of elements.

Probabilistic mathematical Markov processes models are used as methods of information support and decision-making modeling.

4. Methodology

The dynamics of a hierarchical control system can be described using the equations dx dt = f( , , 1, 2, … , ), x(t0)=x0 ( 1 ) where xEm – vector of phase variables, Em – state space at a moment in time t.

The change in the state of information support of the management system occurs under the influence of the control center u(t)U and control subsystems v1(t), v2(t), …, vn(t), vi(t)Vi. Assuming that the control parameter of the center u changes continuously in time, the resulting function u(t), t[t0, t], u(t)U, will be measured by t. Sets U, V1, V2, …., Vn will be the sets of admissible controls.

Every program control u(t), t[t0, t] determines the trajectory of the control system x(t), t[t0, t]. The set of ends of the trajectories of the differential equation ( 1 ) represents the reachability set starting from the initial state for all possible program controls u(t)U, t[t0, t]. Each new event depends only on the previous one and does not depend on all other events. Thus, the resulting control trajectory ends with a point x(t), into which the system passes at time t. This point will be the starting point of the Markov process. The original probability distribution can be represented by the equation: where ∀ – universal quantifier, S – discrete states, q0 – probability distribution at a moment in time t0 = 0.

The set E represents a finite number of possible states.

P( 0 = S) = 0(S) 

E = { 1, 2, … , }

Range of random variable {xn}, the values of which determine the parameters of information support of intelligent control systems, is the state space, and the value n, characterizing the movement of this parameter in the control system, – is the step number. The probabilities of transition from one state to another are represented as square matrices.

(n) = P( +1 = j⃒ = i) ( 2 ) ( 3 ) ( 4 ) P  s2 s1 . sn  p11   p21  .   pn1

. pn2 ... sn ... ... ... ...

p1n 

 p2n  . 

 pnn  Elements, pij denote the probability of transition from the state si into the next.

The transition probability matrix expresses the probability that the state of the control system at time n + 1 is subsequent to other states.

P(xn+1 = n+1⃒xn = n) = P(Sn,Sn+1 )∀(Sn+1 , Sn ) ← E ⨯ E

The Markov chain will be homogeneous if the transition probability matrix does not depend on the step number. ( 5 ) ( 6 ) ( 7 ) ( 8 ) ( 9 )

Pmkj ≜ [Skj⃒Smj-1]

Skj – the probability that the system after j-steps will be in the state Sk, ≜ - mathematical equal sign by definition.

The probability distribution for identifying the state of information support of the control system does not depend on time, but only on transitions from the current state to the corresponding control iterations. By developing the proposed methodological approach, it is possible to establish a sequence of transitions from the initial current state, creating the necessary control base. The stochastic model compiled in this way is shown in Figure 1.

Pij(n) = Pij ( = ⃒ 0 = 0) =

According to the Kolmogorov-Chapman equation, the transition probability matrix for n steps in a homogeneous Markov chain is the n-th power of the transition probability matrix for one step.

Markov chain at any moment of time can be characterized by vectors by a row Ci of the matrix of transition probabilities P.

Transition probability or conditional probability of an event Skj, upon condition Smj-1 is equal to Defining parameters and zones

Determining the information support trajectory

Determining the characteristics of the influence of the external environment

Compilation of the matrix of transition probabilities

Determination of the current state

of information support parameters

Selection of random variables indexed by time

Formation of an information situation related to the reaction to external influences

Establishment of transition state probabilities

Defining discrete state spaces

Construction of a

directed graph

Simulation of random events Construction of stochastic models

efficiency

weight contribution,%

The state of the control system can be described by a random process (t). Random process (t) will be Markovian if its conditional probability distribution function at a future moment of time tn+1 does not depend on the values of the process in the past moments t1, …, tn-1, and is determined only by the value (tn)=xn at the present time tn’. Conditional distribution function P{(tn+1)<xn+1|(tn)=xn}=F(xn, tn; xn+1, tn+1) will be a Markov transition function.

5. Experiment

Below is an example of using information support for intelligent control systems for complex organizational and technical objects based on Markov chains, compiled on the basis of an expert assessment of the management activities of a real production facility (Table 1). certain probability, which is written as a line of the state matrix. In this case, the sum of the probabilities in the rows of the matrix is always equal to one.

∑ =0 [Skj]= ∑

=0 k(j)=1, j≥0

The values of conditional probabilities of information support parameters at different stages of management are presented in Table 2.

Conditional probabilities of information support parameters of intelligent control systems The initial state vector, according to Table. 1 will be written in the form:

The transition probability matrix is the following: p(0)=(0.27, 0.19, 0.17, 0.12, 0.25)

A=||0.19 0.11 the parameters vi, presented in table 1.

The state of information support at the first stage of management S1 will be determined by the first row vector of the matrix A. The probability of the influence of this parameter p( 1 ), characterizing the information security of the management process vi according to the methodology for calculating

Markov chains will be equal to:

The probability that, being in a state S1, information support of the control system will go into the state S2, characterized by unstable, albeit predictable, changes in the environment is p( 2 ). Information support at this stage of management will be determined by the parameter v2. The probability of the influence of this parameter characterizing the security of the control system v2 and its information support from fluctuations in the external environment is equal to:

Comparison of identical parameters presented in ( 14 ) and ( 15 ) showed an increase for the parameters v1 and v3. All other parameters have been reduced. This was the basis for the transition to the next stage of management information support. It should be noted that the state p(0) characterizes only the initial state of the system without control. The first control step starts with p( 1 ). In this case, the probability of the influence of information support parameters at each control step should decrease.

The probability of transition of information support of the control system from the state S3 into a state S4 is equal to p( 4 ). Information support at this stage of management will be determined by the parameter v4. It is determined by cyclic multiplication ( 12 ) by the corresponding row of the transition matrix A.

Comparison of probability parameter values represented by a row vector p(0), with the corresponding probability distributions of the information support of the control system at the first stage of control p( 1 ) ed to the formation of a number of practical recommendations. A comparison of identical values of the parameters presented in ( 11 ) and ( 13 ) showed their increase at the first stage of management information support, which can be considered satisfactory, except for the last parameter "effectiveness" v5, which has decreased. This forces us to move to the next stage of control information support with the probability p( 2 ).

In conditions of instability of functioning due to unpredictable pressures of the external environment, the probability of transition of information support of the control system from the state S2 and S3 is equal to p( 3 ). Information support at this stage of management will be determined by the parameter v3. The probability of the influence of this control information support parameter characterizing controllability v3, that is, the possibility of adjusting control actions is equal to:

Comparison of identical parameters of information support of the control system, presented in ( 15 ) and ( 16 ), showed a decrease in all parameters except v3. This served as the basis for the transition to the next stage of management. The probability of transition of information support of the control system from the state S4 into a state S5 is equal to p( 5 ).

Comparison of identical parameters of management information support, presented in ( 16 ) and ( 17 ), showed their decrease in all parameters. This indicates the quality of information support of the management system.

Oriented graph of Markov chains for the considered example of information support of intelligent control systems for complex organizational and technical objects presents on Figure 1. ( 16 ) ( 17 )

6. Results and Discussion

Markov processes are a convenient mathematical model for various technical applications. When managing complex multi-level organizational and technical objects, depending on the values of a random process (t) and sets of observation moments Markov processes can be divided into moments with discrete and continuous time. The transition from one state of the system to another occurs at regular intervals, which are steps.

Problem situations arising in the process of managing complex multi-level organizational and technical objects, caused by the influence of the external environment, can be methodologically reduced to separate associative homogeneous classes of management decisions, i.e. to some set of strategic alternatives. If the analyzed problem situation is not associated with the existing classes of control alternatives, the next hierarchy level procedure for the formation of control information support is formed, which allows other properties of control information support to be involved in solving the problem.

At the initial moment of time, focusing on the initial conditions, the optimal control of the organizational and technical system is chosen in the time interval [t0, t]. After some time, the state of the system changes, additional information appears, and there is a transition to the next level of the hierarchy of information support for the management process. All possible states of information support process parameters are enumerated. These iterations are carried out for various states of the intelligent control system with their own parameters and probabilities. Since the processes of transformational transformations in the control system do not have a continuous time stamp, a sequence of states can be used as time.

raw data extraction use of additional information parameter input formation of a data array for analysis intuition and expert experience application

software functioning of software

modules No assessment of implementation opportunities

Yes Yes definition of restrictions building a directed

graph evaluation of alternatives output of results end

No

The proposed modernization of the use of Markov chains consists in replacing equal step intervals with a discrete sequence of states, which is determined by unpredictable external environmental influences. Such modernization of Markov chains is a novelty of use in decision-making information support in the management of multi-level organizational and technical objects.

The processes of developing and improving the management of organizational and technical systems operating under conditions of unpredictable influence of the external environment are determined by the information space.

The processes of developing and improving the management of organizational and technical systems operating under conditions of unpredictable influence of the external environment are determined by the information space, the structuring of which is shown in Figure 3.

The performed structuring of the information space for interactive support under conditions of uncertainty makes it possible to assess the scope of information support for intelligent control systems for complex multi-level organizational and technical objects, which is based on software and tools, expert systems, intelligent components, systematization of organizational processes. It is based on information retrieval, data mining, knowledge search in databases, reasoning based on precedents, simulation modeling, neural networks, cognitive modeling. 7. Conclusions 1. For information support of decision making in the management of multilevel organizational and technical objects under conditions of uncertainty, it is proposed to use probabilistic mathematical models based on Markov processes. This makes it possible to simulate and regulate the process of making relevant decisions in real time under conditions of uncertainty. 2. In contrast to the traditional use of Markov chains, information support for decision-making in the management of multi-level organizational and technical objects under the uncertainty of the influence of environmental factors provides for the replacement of equal step intervals by a discrete sequence of states determined by unpredictable external environmental influences in the study of control operations. This will contribute to solving the urgent problem of increasing the efficiency of managing organizational and technical facilities in difficult operating conditions.

8. References

[24] X. Duan, M. George, F. Bullo, Markov Chains with Maximum Return Time Entropy for Robotic

Surveillance (2020) IEEE Transactions on Automatic Control, 65 ( 1 ), art. no. 8675541, pp. 72-86. [25] K.K. Wu, Y. Yam, H. Meng, M. Mesbahi, Parallel probabilistic swarm guidance by exploiting Kronecker product structures in discrete-time Markov chains. (2017) Proceedings of the American Control Conference, art. no. 7962977, pp. 346-351. [26] M. Ficco, Detecting IoT malware by Markov chain behavioral models (2019) Proceedings - 2019 IEEE International Conference on Cloud Engineering, IC2E 2019, art. no. 8790169, pp. 229-234. [27] M. Momenzadeh, M. Sehhati, H. Rabbani, A novel feature selection method for microarray data classification based on hidden Markov model (2019) Journal of Biomedical Informatics, 95, art. no. 103213. [28] T. Pesch, S. Schröders, H.J. Allelein, J.F. Hake, A new Markov-chain-related statistical approach for modelling synthetic wind power time series (2015) New Journal of Physics, 17, art. no. 055001.

References

[1] K. Vorobyova , The Effect of Brand Perception in Malaysia's International Airline Industry During Covid 19 , Annals of Social Sciences & Management Studies , Vol. 6 ( 4 ), 2021 : 555693. doi: 10 .19080/ASM. 2021 . 06 .555693.

[2] O. Bilovodska , A. Kholostenko , Z. Mandrychenko , O. Volokitenko , Innovation Management of Enterprises: Legal Provision and Analytical Tools for Evaluating Business Strategies , Journal of Optimization in Industrial Engineering , 14 (Special Issue), 2021 , pp. 71 - 78 . doi: 10 .22094/joie. 2020 . 677820 .

[3] R. Ludwig , B. Pouymayou , P. Balermpas . et al., A hidden Markov model for lymphatic tumor progression in the head and neck . Sci Rep 11 , 12261 ( 2021 ). Doi: 10 .1038/s41598-021-91544-1.

[4] M. Sharko , N. Shpak , O. Gonchar , K. Vorobyova , O. Lepokhina and J. Burenko , Methodological Basis of Causal Forecasting of the Economic Systems Development Management Processes Under the Uncertainty , Advances in Intelligent Systems and Computing , 2020 , vol. 1246 , pp. 423 - 436 .

[5] T. V. Kozulya , N. V. Sharonova , M. M. Kozulya , & Y. V. Svyatkin , Knowledge-oriented database formation for determination of complex method for quality identification of compound systems . Eastern-European Journal of Enterprise Technologies , 1 ( 2 ( 79 ), 2016 , С. 13 - 21 . doi: 10 .15587/ 1729 - 4061 . 2016 . 60590 .

[6] N. Khairova & N. Sharonova, Building of the logic network of the information area of the corporation , East-West Design & Test Symposium , St. Petersburg, Russia, 2010 , pp. 371 - 373 . doi: 10 .1109/EWDTS. 2010 . 5742044 .

[7] V. Gvozdev , K. Kirillov , Information Support for the Management of the Efficiency of Enterprise Information Service Systems , Proceedings of the 7th Scientific Conference on Information Technologies for Intelligent Decision Making Support (ITIDS 2019 ), Vol. 166 . doi: 10 .2991/itids19. 2019 . 5 .

[8] M. Sharko , N. Gusarina , N. Petrushenko , Information-Entropy Model of Making Management Decisions in the Economic Development of the Enterprises . In: Lytvynenko V., Babichev S. , Wójcik W. , Vynokurova O. , Vyshemyrskaya S. , Radetskaya S . (eds) Lecture Notes in Computational Intelligence and Decision Making. ISDMCI 2019. Advances in Intelligent Systems and Computing , 2020 , vol 1020 . Springer, Cham, pp. 304 - 314 . doi: 10 .1007/978-3- 030 -26474- 1_ 22 .

[9] V. Babenko , O. Nakisko & I. Mykolenko , Research of the aspects of modeling of the project management of risk of implementation system information support , Technology Audit and Production Reserves , 1 ( 4 ( 39 ), 2017 , pp. 64 - 69 . doi: 10 .15587/ 2312 - 8372 . 2018 . 124538 .

[10] B. Michalik , M. Keutel and W. Mellis , Coping with Requirements Uncertainty -- A Case Study of an Enterprise-Wide Record Management System," in 2014 47th Hawaii International Conference on System Sciences (HICSS) , Waikoloa, HI , USA, 2014 , pp. 4024 - 4033 . doi: 10 .1109/HICSS. 2014 . 498 .

[11] T. Lechler , T. Gao , D. Keeney , & B.H. Edington , Exploring contextual conditions of project uncertainties and project value opportunities . Paper presented at PMI® Research and Education Conference , Limerick, Munster, Ireland. Newtown Square, PA: Project Management Institute , 2012 .

[12] M.A. Hasani , M. Regan , Understanding Risk and Uncertainty Management Practice in Complex Projects , European Scientific Institute, edition Vol. 4 , No. 4 , December 2017 , pp. 24 - 38 . doi: 10 .19044/elp.v4no4a3.

[13] M. Sharko , I. Lopushynskyi, N. Petrushenko , O. Zaitseva , V. Kliutsevskyi , Y. Yarchenko , Management of Tourists' Enterprises Adaptation Strategies for Identifying and Predicting Multidimensional Non-stationary Data Flows in the Case of Uncertainties . In: Babichev S., Lytvynenko V. , Wójcik W. , Vyshemyrskaya S . (eds) Lecture Notes in Computational Intelligence and Decision Making. ISDMCI 2020. Advances in Intelligent Systems and Computing , 2020 , vol 1246 . Springer, Cham, pp. 135 - 150 . doi: 10 .1007/978-3- 030 -54215- 3 _ 9 .

[14] L. Dong , D. Neufeld and C. Higgins , Top Management Support of Enterprise Systems Implementations, Journal of Information Technology , 24 ( 1 ), 2009 , pp. 55 - 80 . doi: 10 .1057/jit. 2008 . 21 .

[15] D. Lipaj & V. Davidavičienė , Influence of Information Systems on Business Performance , Mokslas - Lietuvos ateitis, 2013 . doi: 10 .3846/mla. 2013 . 06 .

[16] M.V. Sharko , A.V. Sharko , Innovative aspects of management of development of enterprises of regional tourism , Actual problems of economy , 7 ( 181 ), 2016 , pp. 206 - 213 .

[17] M. Sharko , O. Gonchar , M. Tkach , A. Polishchuk , N. Vasylenko , M. Mosin & N. Petrushenko , Intellectual Information Technologies of the Resources Management in Conditions of Unstable External Environment , International Scientific Conference “Intellectual Systems of Decision Making and Problem of Computational Intelligence” , 2021 , pp. 519 - 533 . doi: 10 .1007/978-3- 030 - 82014-5_ 35 .

[18] H.G. Gómez , M.D.A. Serna , R.F.O. Badenes , Evolution and trends of information systems for business management: the m-business. A review, DYNA , Vol. 77 , 2010 , pp. 181 - 193 .

[19] A. Lesko , I. Prychep & T. Lesko, Development of approach to anticipatory risk management of the enterprise under uncertainty conditions , Technology Audit and Production Reserves , 4 ( 4 ( 36 ), 2017 , pp. 9 - 15 . doi: 10 .15587/ 2312 - 8372 . 2017 . 108595 .

[20] K. Goher , E. Shehab , A-A. Ahmed , S. Shoaib , Towards an Uncertainty Management Framework for Model-Based Definition and Enterprise EasyChair Preprint no . 6807 , 2021 doi:10.3233/ATDE210091.

[21] P. Bidyuk , Y. Matsuki , A. Gozhyj , V. Beglytsia , I. Kalinina Features of application of monte carlo method with Markov chain algorithms in bayesian data analysis . Advances in Intelligent Systems and Computing , 1080 AISC, 2020 , pp. 361 - 376 .

[22] D. Zang , J. Liu , H. Wang , Markov chain-based feature extraction for anomaly detection in time series and its industrial application (2018) Proceedings of the 30th Chinese Control and Decision Conference , CCDC 2018 , pp. 1059 - 1063 .

[23] J. Balak , K. Rastocny , Use of tensor construction of Markov chains when evaluating observed feature of E-SRS ( 2018 ) 12th International Conference ELEKTRO 2018 , 2018 ELEKTRO Conference Proceedings , pp. 1 - 6 .