Predictive Analytics In Single-Industry Town Risk Management

Abstract

The relevance of using predictive analytics tools in the process of forecasting and analyzing risks in the field of solving problems of single-industry towns is due to the fact that the existing arsenal of identifying these threats does not always allow us to notice negative trends in time besides, no single-industry towns support programs were implemented during the period of 2016-2018 in the Russian Federation. The results of ranking the risks and development factors of territories of socio-economic development (TOSED) in single-industry towns are presented (on the example of the city of Yurga, Kemerovo region). The effectiveness of the use of predicative analytics in the risk management of TOSED of single-industry towns is substantiated. The types of classification, the possibility of their use in identifying threats to the development of a single-industry town are considered. The sources of data collection and accumulation of data for forecasting threats to the creation and development of TOSED in the single-industry towns of the Russian Federation are identified. Recommendations are offered to potential residents of TOSED and municipal authorities of single-industry towns with such status on forecasting the risks of creating and progress of TOSED. The obtained results contribute to the development of risk management, project management in the field of regional and municipal economies.

Keywords: Developmentpredicative analyticsrisksingle-industry townterritory of priority social and economic development

Introduction

The early completion of the priority program “Integrated Development of Single-industry Towns” (“Report on the results of the expert-analytical event …”, 2019) due to the lack of significant results is largely since the risks of failure of program activities and the factors of their low efficiency were not predicted. As a result, the problems of many single-industry territories of the Russian Federation have not been resolved, entrepreneurs have not been able to implement their investment projects, TOSER, created in single-industry towns have not received an influx of residents and investments. Obvious and applicable threat indicators did not ensure timely detection of negative trends. For example, in the city of Yurga, Kemerovo Region, LLC “United Woodworking Commercial and Industrial Company” planned to invest 626.6 million rubles in the creation and development of their business at TOSER “Yurga”. However, the agreement was terminated. A TOSER resident from Tomsk, “Mir” company, planned to build a propolystyrene concrete production plant at TOSER of Anzhero-Sudzhensk, Kemerovo Region, worth about 100 million rubles. But excessive interference of the TOSER resident in the business through tax audits and the imposition of interim measures on the company’s accounts led to the loss of this investor. Obviously, in this case, traditional risk management tools are not effective both for enterprises - potential investors and residents of TOSER, and for authorities. It makes sense to look for patterns in the phenomena and processes taking place in the territories of single-industry towns not only with the help of well-known classification features and risk factors. Otherwise, the decision currently being discussed on the integration of all measures to support single-industry towns into one priority national project will not bring the desired benefit either to the state, or to entrepreneurs, or to the population of single-industry towns. In the meantime, the lack of innovation in the field of risk forecasting has led to the fact that the amount of tax revenue decreased in 2018 in 70 single-industry towns of the Russian Federation, the number of residents of all single-industry towns decreased by 38.6 thousand people in the period of 2016-2018, and the unemployment rate increased and in some cases exceeded the average Russian rate by 3 times (“Report on the results of the expert-analytical event …”, 2019). In this regard, the integration of modern forecasting tools and risk management of TOSER monotowns, as well as potential residents and investors of such territories, becomes urgent.

Problem Statement

The traditional arsenal of risk management methodology in modern conditions does not cope with the tasks of finding effective solutions in the field of regional and municipal management. Foreign and Russian companies in various industries are beginning to use predictive analytics actively, in particular, to reduce risks, increase business efficiency, marketing, and logistics. However, such a methodology has not yet been used to anticipate trends in the development of single-industry towns (Ivanova, Trifonov, & Nesteruk, 2019). The adaptation of predicative analytics tools to the tasks of managing the development of single-industry towns, in particular in the field of development risk management, is in demand. The authors solve the problem of selecting predictors that make it possible to make high-quality forecasts of changes in the situation in single-industry settlements depending on the support measures applied, for example, obtaining TOSER status, the behavior and interests of investors, potential residents of such territories, city residents, and authorities. The objects of the research presented in this article were the problems of single-industry towns, the characteristics and features of TOSER in single-industry towns, affecting the sustainable development of TOSER and the activities of residents in such territories. The research information base was research of the practice of solving the problems of single-industry towns, officially published statistical data relating to the establishment and operation of TOSER in the Russian Federation, the functioning of single-industry towns, the implementation of various instruments to diversify their economies, empirical data collected through questionnaires, surveys and interviews of real and potential residents of TOSER single-industry towns of Kemerovo region.

Research Questions

The article presents the results of a classification (a type of predictive analytics) that studies both predictor variables and classifier response variables that allow predicting risk events in the trajectory of creating and developing TOSER single-industry towns.

Purpose of the Study

The authors chose the selection of predictors and risk-classifying variables that ensure the prediction of risk situations in the decision-making process on creating a TOSER single-industry town, developing and implementing a TOSER development program, interacting with investors, potential residents of the territory, as well as in the process of developing and implementing investment projects entrepreneurs in single-industry settlements as a goal of the study.

Research Methods

Findings

6.1. Recommendations on the use of predicative analytics tools in risk management for creating and developing TOSER of single-industry towns

In the field of developing programs for creating and developing TOSER, it is necessary not only to introduce a culture of leadership in the processes of socio-economic transformation, but also tools for predicative analytics to ensure the validity of preventive anti-risk measures. Khasanov (2018) identifies types of predictive analytics: forecasting the present and shaping the future. In the first case, regularities are determined in the present tense, under existing conditions. In the second case, the identification of a new paradigm as a result of the accumulation of data on phenomena that are not typical for today takes place. Therefore, predicative analytics can not only predict, but also create data arrays and influence the situation in the future. A different methodology for collecting information is currently required for forecasting trends. Assumptions and scenarios should highlight risk and uncertainty. At the same time, it is important not only to formulate and define realistic assumptions, but also to use indicators that are subject to control, as well as uncontrolled variables that can only be tracked. In this regard, we propose to form data arrays that include not only statistical indicators generated by state statistics bodies in the process of making and implementing decisions on the development of single-industry towns. In many cases, data collected (or available), for example, by the tax inspectorate, police, banks, insurance companies, carriers, etc. can be used as predictor variables.

To anticipate risky events during the implementation of projects to create TOSER in single-industry towns, we propose the use of classification. Classification (Gutierrez, 2017) as a type of predictive analytics involves the use of answer options, divided into several categories (large, medium, small, for example). The classifier studies a set of indicators, including information about the response variable and predictor variables. With the help of staging (algorithm), combinations of variables associated with the answer are revealed. The set of indicators is a training set. Next, the algorithm explores new observations in which there is no information about the desired answer. The algorithm assigns a classification to new observations, based on the classification of the training set of indicators.

Classification methods are the method of nearest neighbors, decision trees that support vector machines, the naive Bayesian classifier, random forests. For large samples, approximate methods for finding nearest neighbors can be used. It should be borne in mind that the quality of the classification of nearest neighbors is determined by the number of neighbors, the metric of the distance between objects (Hamming metric, Euclidean distance, for example), the weight of neighbors. It is recognized that the method of nearest neighbors is simple to implement, well studied theoretically, amenable to interpretation. However, if there are a lot of features in the data set, this makes it difficult to select weights and cut off unimportant features. Decision trees are easily visualized, require a small number of model parameters, it is realistic to use both numerical and categorical features. However, the model allows only interpolation, but not extrapolation. The “random forest” method (Chistiakov, 2013) is based on the construction of many decision trees, each of which is constructed according to a sample formed from the initial training sample. In contrast to the classical algorithms for constructing decision trees (Breiman, Friedman, Olshen, & Stone, 1984; Quinlan, 1987), this method assumes that when constructing each tree at the vertex splitting points, a fixed number of randomly selected features of the training set are used. Это означает, что каждый лист дерева включает наблюдения исключительно одного класса. The naive Bayesian classifier (a set of classification algorithms that accept the assumption that each parameter of the classified variables is examined independently of other class parameters) can be quite accurate and can be used to filter in the field of predicting risk phenomena.

We determined the indicators and predictors proposed for the answer to predict the risks of implementation and achievement of the planned indicators of projects for the creation of TOSER in single-industry towns (see Table 02 ).

The presented predictors can be used in the classification to predict risks, not only if the planned indicators of the TOSER program are not fulfilled in the single-industry town, but also when risks are foreseen by potential residents in the decision-making process on the location of the enterprise in the city with a special status.

Conclusion

The presented results allow not only to form managerial decisions in the field of risk management for the development of TOSER of single-industry towns, but also develop the existing methodology for forecasting risks in the field of municipal and regional management.

Acknowledgments

This article was prepared with the support of the Russian Federal Property Fund in the framework of the project “Development of methodological tools for managing single-industry towns of the Kemerovo region using predicative analytics to counteract the sources of danger of society and the economy of the state” No. 18-010-00244 A, 2018 competition of fundamental research projects, number TsITiS AAAA- A18-118011990067-1 (Trifonov V.A.).

References

1. Breiman, L., Friedman, R., Olshen, R., & Stone, C. (1984). Classification and Regression Trees. Belmont, California: Wadsworth International.
2. Broder, J. F., & Tucker, E. (2006). Risk Analysis and the Security Survey. GB: Butterworth-Heinemann.
3. Chichkanov, V. P., & Belyaevskaya-Plotnik, L. A. (2016). Analysis of the Approaches to the Assessment of Regional Processes of Formation of Social and Economic Security. Ekonomika regiona, 12, 654-669.
4. Chistiakov, S. P. (2013). Random forests: an overview. Proceedings of the Karelian Scientific Center of the Russian Academy of Sciences, 1, 117–136.
5. Glazyev, S. Yu. (1997). Basis of ensuring the economic security of the country: the alternative reformation course. Rossiyskiy ekonomicheskiy zhurnal, 1, 8-9. [in Russ.].
6. Gutierrez, D. D. (2017). InsideBIGDATA. Predictive Analytics Guide. Retrieved from http://www.spotfiretibco.ru/wp-content/uploads/2017/09/InsideBIGDATA.pdf
7. Illarionov, A. N. (1998). Kriterii ekonomicheskoy bezopasnosti. Voprosy ekonomiki, 10, 35-58. [in Russ.].
8. Ivanova, O. P., Trifonov, V. A., & Nesteruk, D. N. (2019). Directions and possibilities of predictive analytics in managing the development of single-industry towns. Espacios, 40(3), 04.
9. Kachalov, R. M. (2012). Management of economic risk: theoretical bases and applications. Saint-Petersburg: Nestor-Istoriia. [in Russ.].
10. Kachalov, R. M. (2018). Methodical Issues in Assessing the Risk of Failure to Achieve the Strategy Goal of the Monotown Development. Bulletin of Kemerovo State University. Series: Political, Sociological and Economic Sciences, 1, 97-102. https://doi.org/10.21603/2500-3372-2018-1-97-102. [in Russ.].
11. Kachalov, R. M., & Sleptsova, Iu. A. (2017). The analysis of risk in activity of the enterprise from a position of the system economic theory of Kleyner G. B. Ekonomika i upravlenie: problemy, resheniya, 8, 66-70. [in Russ.].
12. Khasanov, A. R. (2018). Impact of predictive analytics on the activities of companies. Strategic decisions and risk management, 3, 108-113. https://doi.org/10.17747/2078-8886-2018-3-108-113 [in Russ.].
13. Kleiner, G. B. (2013). System economy as platform of development of the modern economic theory. Voprosy ekonomiki, 6, 4-28. [in Russ.].
14. Kleiner, G. B. (2015). System resource of strategic stability of economy. St. Petersburg state polytechnical university journal. Economics, 4, 10–24. [in Russ.].
15. Ogorodnikov, P. I., Zaloznaya, G. M., & Borovsky, A. S. (2018). The System Analysis of Ensuring the Stability of Innovative and Digital Economy on the Basis of Intellectual Comprehensive Security System. Economy of Region, 14(4), 1221-1231. [in Russ.].
16. Quinlan, J. R. (1987). Simplifying decision trees. International Journal of ManMachine Studies, 27, 221–234.
17. Report on the results of the expert-analytical event “Monitoring and evaluation of the implementation of the priority program “Integrated development of single-industry towns”. (2019). Retrieved from http://www.ach.gov.ru/activities/bulleten/964/37982
18. Senchagov, V. K. (2002).  Economic security: geopolitics, globalization, self-preservation and development: a textbook.  Moscow: Finstatinform Publ. [in Russ.].
19. Tarasova, O. V., & Rudneva, V. A. (2017). Models of revitalization of old industrial territories. IVF, 7, 1.