Текстовые фрагменты публикации
ББК 65.291.2УДК 338.24
Моделирование ситуационных стратегий с применением нейронных сетей на примере промышленного предприятия
Modeling of situational strategies with application of neural networks on the example of the industrial enterprise
Volovikov Boris
Omsk Institute (branch) of Russian State Trade Economic University
The summary
In the article a new approach to the choice of strategy of the enterprise’s development, based on the use of neural networks is given. The elaborated hybrid algorithm has been approved in the process of working out the strategic plan of the industrial enterprise’s development, the adequacy is estimated. The advantages of the method and its characteristic features are revealed.
Keywords
Enterprise’s strategies, neural networks, method of decision tree, situational analysis
Strategic management as a scientific trend is constantly developing, improving its methodological base. The vector of evolutional development of methods of the industrial enterprise’s development management is directed to a fuller use of information and reasonability of the approach for making the managerial decisions. The peculiarities of modern methodology of the enterprise’s strategic management are, on the one hand, the factor of uncertainty, caused by processes’ stochasticity in the external environment, on the other - the necessity of making concrete managerial decisions. Explicit contradictions between the presence of the conditions of uncertainty in the external environment and the necessity to make a definite managerial decision, demanding the quantitative assessment, create the necessary prerequisites for creating a new approach to forming the enterprise’s strategy. There is a demand in the approach that will allow uniting methods of external environment investigation under conditions of uncertainty and methods of strategic planning.
At the present moment trends, related to application of artificial intelligence systems, are successfully developing in Economics and Sociology. On the basis of collected information and experts’ experience, artificial intelligence systems allow providing the support to making managerial decisions on the basis of quantitative assessment of external environment. Well-known methods of matrix analysis don’t allow taking into consideration those regularities of external and internal processes, and also all the factors, that allow analyzing the situation on the
market and within the enterprise fully and thoroughly. Possessing a rigid logic of making decisions, these methods do not allow making forecasts in the dynamics of changing the enterprise’s strategic course of development. With characteristic simplicity and visualization, they simplify a real state of things, suggesting banal decisions, known by the director in advance [1,2]. A difficult task for matrix analysis is also taking into account the significance of all particular criteria, determining key index. The method of expert evaluation, suggested for determining weighting coefficients, complicates using these methods more from the point of view of organization of this process, as well as from the point of view of trustworthiness of the results. Also two-dimensional graphic presentation of the results of matrix analysis, for example, while carrying out the portfolio analysis, distorts the picture of strategic positioning of business units.
In the article a new approach to the elaboration of the enterprise’s strategic development, based on the applying the method of neural network, is suggested. The goal of the research is to create a new methodology of forming a package of the industrial enterprise’s alternative strategies. A new methodology implies creating a hybrid algorithm, that includes the sequence of application of two methodologies, one of which is intended to form the portfolio of more effective strategies on the basis of neural networks’ method, the second one is meant to evaluate each factor on the basis of the decision tree method while forming the situational strategy. At work the notion of situational strategy is used, that is formed on the basis of analysis of the situation in the external environment and the enterprise’s condition. Unlike the well-known method of SWAT-analysis, in the result it is supposed to get not a recommendation for a rational choice of combination of opportunities, threats, strong and weak points, but a concrete set of business-strategies. The adequacy of the suggested hybrid algorithm was estimated on the example of elaboration of the concrete enterprise’s strategic plan of development
In order to achieve the set goal the following tasks were stated and solved:
1. Development of methodology of forming the enterprise’s alternative situational strategies, which are the most suitable to the external environment conditions and the enterprise’s resource potential taking into account indeterminacy and stochasticity of the processes.
2. Control of logics of the algorithm of the choice of strategies with application of decision tree method and estimation of the significance of each factor, having influence on the choice of strategy.
3. The choice of strategic priorities of the industrial enterprise’s development on the practical example with application of the above-listed methods.
2
The industrial enterprise’s strategic planning and resources management activity was chosen as the subject of research.
The novelty of the research is taking into consideration those environmental factors and regularities, which cannot be simultaneously taken into account by means of all known methods environmental analysis, and also the opportunity of managing the logics of the strategic choice not by means of subjective evaluation of the author’s strategy, but by means of monitoring of the real situation.
On the whole, the suggested methodology allows formalizing the procedure of the choice of strategic alternatives and elaborating the program product that automatizes the strategic planning
process.
In the picture 1 there is a block-scheme of algorithm of the choice of strategy that is the most suitable to the enterprise’s goal
Pic 1. The block-scheme of algorithm of the choice of strategy
3
Each of the above-listed stages includes the sequence of logic operations, connected with each other by means of reverse connections. On the first stage the factors, determining the conditions of the strategy’s choice, and the limits of their change were determined. In the table 1 the names of the indices, their definition and limits of variations of banal meanings were given as an example.
Table 1. The indices of the strategy’s choice ¹
Name of Range of indexes’ changes1 2
the index Definition 1- the worst 2 -average 3 - beneficial
meaning
Characterizes the
market market’s condition Market reduction Stable condition Market growth
in general
Is characterized by Average High
product the index to the Low competitiveness competitiveness
product’s competitiveness
competitiveness
Characterizes the Unsatisfactory
suppliers level of efficiency Satisfactory Good
of the delivery
system
competit The level of threat 0 - the absence of threat; 1- the high level ol f competitors’ activity
ors from the
competitors
marketin Marketing Unsatisfactory Satisfactory Well-developed
g organization marketing system
Absence of Sufficiency of Sufficiency of
financial financial resources financial resources resources for the
resources for the strategy for the realization of realization of
realization one product several products
producti Condition of the Insufficiency of Average condition Sufficiency of
on resource production capacity of production base production capacity
resources for the strategy for realization of the
realization whole portfolio
innovativ Presence of Presence of the
e Absence of innovations for the balanced innovative
resources innovations strategy realization portfolio
¹ Depending on the problem definition, some changes in the indexes’ structure are possible
² —zero” value of index is possible, that means absence of the factor’s influence on the situation
4
The choice of indices’ values can be made by means of the method of expert evaluations, though it is more preferable to use the results of marketing research of external environment and business diagnosis of the enterprise’s condition. In each concrete case it is recommended not to choose only characteristics of input, which are the most significant from the point of view of the researcher, but also to determine the range of their values’ changes. In particular, for the example given further, the predicted sales volume and market trend were used as the index —market”, —competitive capacity” was determined by means of calculations of values of the complex index of competitiveness of all the market players’ products. For the financial sphere the presence of sufficient investment volume is the priority question, for industrial sphere - the value of production capacity, for innovative - the presence of innovations, which allow maintaining the product’s competitiveness and strengthening the competitive positions of the enterprise. The indices —suppliers” and —marketing” can be characterized accordingly, by means of analysis of diagram of components’ deliveries and information about sales during the former periods. The degree of competitive intensity can be measured by Hirshman index or by means of analysis of the competitors’ trend sales (the last option is preferable). The question concerning determining resource-intensity values is arguable. While choosing a strategy for each enterprise it is necessary to elaborate evaluation system for the degree of resources’ availability for the strategy realization.
Any combination of values of the above-listed input characteristics corresponds to several alternative variants of a strategy. In the table 2 individual examples of choosing businessstrategies according to the values of environmental input characteristics are given.
Table 2 Evaluation of indices of business-strategies.
Name of a business- Indices’ values, determining the choice of strategy
strategy Market Product Supplie Comp Mark Resources
rs etitors eting Finan Industr Innov.
Direct integration 2 3 3 1 1 3 0 0
Reverse integration 2 3 1 1 3 3 0 0
Horizontal 2 3 1 1 3 3 0 0
integration
Seizure of the 3 3 2 0 3 3 0 0
market
Market development 3 3 2 0 3 3 3 0
Product’s 2 1 2 0 3 3 3 3
development
Concentric 3 3 2 0 2 3 3 3
diversification
Conglomerate 1 3 2 0 2 3 3 3
diversification
Horizontal 2 1 2 0 2 3 3 3
5
diversification
Strategic alliance 3 3 1 0 2 1 1 3
Reduction 1 1 1 1 1 1 1 1
Restructuring 0 0 1 1 0 1 0 0
Enterprise’s 1 1 1 0 1 1 1 1
liquidation
The examples, given in the table 2, are taken from the concrete situations from the history of the industrial enterprise’s development. Under conditions of uncertainty it is difficult to find an ideal situation for the choice of strategy, usually several alternative strategies can be suitable for a concrete situation. Besides, value variation is possible in estimation of any input characteristic due to expert subjectivity. Insignificant deviations because of stochasticity and randomness of the environmental processes can change values of input characteristics’ evaluations, i.e. conditions for the choice of the same strategy for different values of input characteristics, are created. The complex of external and internal factors create a weak-structured environment, modeling of which demands application of specific methods of research.
The problem of influence of stochasticity and randomness of the processes, and also the subjectivity of expert evaluations allows solving methods of neural networks and the method of decision tree. Advantages and disadvantages of these methods were discussed very often in scientific literature [3,5,6], but in this article the attention is focused on the results, got while modeling the process of choosing a strategy with application of these methods. Under conditions of the processes’ stochasticity in the external environment, neural networks manage to reveal the regularities, which cannot be exposed by other methods. In this situation the matrix method does not allow taking into consideration such quantity of factors, which will allow choosing the set of strategies —manually”. With the increase of quantity of input characteristics and range of evaluations, application of matrix methods makes solving this task impossible.
The important aspect for neural networks is creating educational selection and the choice of the settings’ characteristics. Educational selection conveys the information about the object under research to the neural network, —teaching” all the rules and laws of behavior, which are characteristic for the object under research. The more information the educational selection contains, the more reliable results of the neural networks research will be. In the examined situation of strategic planning the indices of internal and external environment will be an input for the neural networks, the set of strategies recommended for this concrete situation will be an output. The educational selection must contain a complete set of situations and strategic solutions. In the table 2 a part of the educational selection is shown, for the concrete task of
6
choosing a strategy, the volume of information with the described situation and strategies is much more considerable.
The multilayer perseptron (9 inputs, 3 layers, 12 outputs), a common variant of a neural network, was chosen for the research. The model was evaluated by comparing strategies (recommended by the neural network) with actual values, taken from practice. As a practical example we considered the choice of strategies of enterprise’s development - manufacturer of radio connection, the main activity of which is radio relay system of radio connection, mobile connection complexes, and gas meters - the direction that is not core for it. The enterprise has a goal of determining the strategic direction of development for each direction. The primary goal is to form a portfolio of alternative decisions for further concentrating efforts on the most attractive project. In the table 3 for each direction situations are given and estimated appropriately. Neural networks for each situation allow finding situational strategies, which are the most suitable for the given situation.
Table 3. Initial data for the choice of strategy
Number Market New Compiti Suppliers Competitors Sales Resources
of market vita of Financial Industr. Innovat.
situation product
1 1 2 2 2 1 2 0 1 1
2 1 1 2 2 1 3 1 2 1
3 1 1 0 1 1 2 0 1 0
Comparing the results of modeling with the certified strategies, allowed making a conclusion concerning the degree of the model’s credibility. The strategies which did not coincide with the actual values were italicized. In particular cases the difference is not principal, for example, diversification strategies, recommended in case of stagnation of the traditional product market, are similar to each other in the content, because they reflect the general direction - the diversity of nomenclature of the produced products. It is better not to make a conclusion about considerable divergence between the research results and actual values. As strategies, alternative for two situations, neural network recommends strategies of reduction (reduction and rejection). It indicates that a neural network, possibly having evaluated the situation, perhaps, more thoroughly, gives a more reserved evaluation, recommending to think about reduction of the production or restructuring. In this case neural network, perhaps, —found out” threats in this situation, which had not been revealed by the marketing service in this organization. An absolute advantage of neural networks is also automation of the whole modeling process that implies the opportunity of increasing the quantity of input and output variables. In case of using not 12 types
7
of strategies, but, for example, 34 stratogems as output characteristics, the choice of alternative strategies is impossible with the help of intuition only.
Table 4. The comparison of the results of modeling with actual values
The most significant
Number of The direction Actual strategies Calculated factors, revealed in
situation of activity strategies modeling process
Concentric and Rejection,
1 Mobile horizontal concentric and
connection diversification, horizontal
complexes strategic alliance diversification
Gas meters Conglomerate Conglomerate
diversification, diversification, 1. Industrial resources
2 horizontal horizontal 2. Efficiency of
diversification diversification, deliveries
reduction 3. Efficiency of sales
Horizontal Reduction,
Radio relay diversification, horizontal
3 stations reduction diversification,
concentric
diversification
Let us stop at practical application of the received results. Each of the suggested strategies can exist, in case there is a concrete opportunity of its realization. In every direction of activity the monitoring of projects was conducted according to the chosen strategies. For example, in the field of gas meters the production overstocking was found out, so there is a demand in urgent actions to decrease self-cost of the product in order to stimulate sales volume. For realizing this strategy as one of the recommended, the program of the meter’s modernization, providing phased decrease in the product’s price and thus, the price of sales. The strategy of horizontal diversification (i.e. elaboration and promotion of new products on the market) is suitable for directions —mobile connection complexes” and —radio relay stations”.
After evaluation of the model’s adequacy the research on the evaluation of the significance of input factors in the modeling process. The method of decision tree allows classifying all the input indices in accordance with features, properly classified in each neural network node, and tracking the logics of the strategy’s choice. Among the most influential strategies are the following ones: the industrial resources condition (the quantity of examples, classified by this tree node is 106), suppliers (this number is 77) and marketing organization (respectively - 74). This fact indicates that industrial resources and organization of logistic processes have a great impact on choosing a strategy. Changes in market factors of that scale, similar to the one of the educational selection, have a less influence.
8
In the conclusion the following most significant aspects of the research were pointed out and the following conclusions were made:
1. Forming a portfolio of situational strategies on the basis of neural networks allows determining a vector of the industrial enterprise’s strategic development under conditions of stochasticity of the processes, which characterize the external environment and analysis of the enterprise’s condition.
2. The procedure of modeling of strategic planning with application of the decision tree method includes the opportunity of quantitative evaluation of each input characteristic’s significance and allows concentrating efforts on the most important enterprise’s business processes from the point of view of the strategy’s realization.
3. The flexibility of neural networks method allows changing the task formulation and taking into consideration various groups of factors, influencing on the strategic choice: social-economic, political, ecologic, legal, etc. The result of the research depends on the degree of the researcher’s awareness and significance of the above-mentioned characteristics for the enterprise’s business.
4. Methodology of neural networks simplifies the procedure of the search of the most correct decision with the help of the automatic choice of approximate functions in each node of neural network and provides the minimum calculative mistake and, consequently, the reliability of the results of the research.
The bibliographic list
1. Volovikov B.P.Application of the complex approach to formation of corporate strategy. Practical marketing N9 (163)-2010, with. 32-36.
2. Volovikov B.P.Elaboration of an optimum business portfolio of the industrial enterprise. The bulletin of OmGU a series "Economy", N1, 2010-s.162-170.
3. Debok Г, Kohonen Т, the Analysis of the financial data with the help of self-organized cards / the Lane with English - М: «Alpina», 2001. With. 19.
4. Lamben Jean-Jacque. The management focused on the market / of Recentury with English Under the editorship of V.B.Kolchanova.-Spb.: Peter, 2006.-800 with.: silt. - (A series «Classics МВА)
5. Levitin A.V. Algorithms: introduction in working out and the analysis.: the lane with English — M:Edition house "Williams", 2006. — 576 with.: silt. — ISBN 5-8459-0987-2
5. Robert Kallan the Basic concepts of neural networks The Essence of Neural Networks First Edition. — 1. — "Williams", 2001. — With. 288. — ISBN 5-
9
Фрагмент текстового слоя документа размещен для индексирующих роботов.
Для полноценной работы с документом, пожалуйста, перейдите в
ридер.
