=Paper=
{{Paper
|id=Vol-2105/10000077
|storemode=property
|title=Machine Learning in Estimating of SME Investment Potential in Ukraine
|pdfUrl=https://ceur-ws.org/Vol-2105/10000077.pdf
|volume=Vol-2105
|authors=Alla Ivashchenko,Yevheniia Polishchuk
|dblpUrl=https://dblp.org/rec/conf/icteri/IvashchenkoP18
}}
==Machine Learning in Estimating of SME Investment Potential in Ukraine==
https://ceur-ws.org/Vol-2105/10000077.pdf
Machine Learning in Estimating of SMEs Investment
Potential in Ukraine
Alla Ivashchenko1, and Yevheniia Polishchuk2
1
Kyiv National Economic University named after Vadim Hetman, Corporate finance and con-
trolling department, Kyiv, Ukraine
2
Kyiv National Economic University named after Vadim Hetman, Investment activity depart-
ment, Kyiv, Ukraine
1alla.ivashchenko@kneu.edu.ua, 2yivga_83@ukr.net
Abstract. The aim of this research is to develop a model of SMEs investment
potential assessment, using some of machine learning approaches. The structure
of investment potential for SMEs was defined underlining their main character-
istic features. It was revealed that the SME investment potential depends on fac-
tors of business environment measured by indicators of Annual Doing Business
Reports developed by World Bank Group. The methodology for assessing the
impact of business environment factors on the SMEs investment potential was
developed. This methodology is based on the algorithm of machine learning,
which can be used to design a model for forecasting the investment potential of
SMEs. This model allows to determine the degree of influence of parameters on
the formation of the SMEs investment potential. It is recommended to use com-
puter language Python for optimization of time and human resources. It provides
the opportunity to study the effects of the main drivers (both enhancement and
reduction) of SMEs investment potential aimed at its improvement. The authors
revealed that estimation results could become the basis for elaboration of recom-
mendations regarding improvement of business environment in Ukraine.
Keywords: Machine learning, SME, Investment Potential, Business factors,
Assessment model, Python
1 Introduction
The level of investment attractiveness plays the pivotal role for business development
of every country. There are different types of ranking methodologies provided by in-
ternational institutions covering various aspects of conditions for running business
which objective results can help to observe the level of financial sustainability either to
invest financial resources into a country`s economy or to abstain from it. Doing Busi-
ness reports include economic indicators which can be used as information tool for
identification of factors negatively or positively influencing investment attractiveness
of economy and economic development on the whole.
Business environment factors as the elements of Doing Business methodology in-
fluence the investment potential of enterprises, small and medium sized (SMEs) in par-
ticular. Investment potential of legal entities defines the level of a separate country eco-
nomic growth and its sustainability. That is why it is essential to assess the mentioned
factors impact for investment potential formation and forecasting it which can be made
by using Machine learning approaches.
In the research the effect of business environment factors on investment potential of
small enterprises was defined by automatic relevance determination regression model.
Besides, for medium ones it is reasonable to use classical linear regression for the men-
tioned purpose.
So, problems of investment potential growth and negative impact of different busi-
ness factors need to be solved in order to minimize the business risk level, to increase
the level of funding for enterprises, e.g. SMEs and to forecast the prospects of invest-
ment potential formation for SMEs.
2 Literature Review
The potential as a scientific category is universal in application. This allows it to be
used in various fields of science: in biology, mathematics, physics, medicine and eco-
nomics (particularly in investment). In economics investment, innovative, personnel,
financial, technical potentials have been described in different literature studies. In this
research investment potential of SMEs has been considered because it determines the
directions of their activity at the investment market. It forms the grounds for further
investment. Investment potential defines the axis, strategy, and performance of SMEs.
Investment potential at the macro level shows potential volumes of attraction of finan-
cial resources to investment processes, revealing the contribution of SMEs to the coun-
try's economic growth. Permanent monitoring and measurement of the investment po-
tential of SMEs allows their investment activity and their investment preferences to be
tracked, and to determine the problems of feasibility of using resources for further in-
vestment.
There is a range of published studies describing the investment potential and ap-
proaches to its estimation, but in some scientific papers we can meet the term “invest-
ment opportunities”. In this study, we consider investment potential and investment
opportunities as related categories. Previous research reflects the approaches which can
be divided into several groups related to: (1) the resource, (2) market, (3) probabilistic,
(4) resultative, (5) capacitive, (6) structural, (7) cost.
S. Leonov [1], J. Eklund [2], N. Yaremchuk [3] & P. Dieterlen [4] are proponents of
the first group related to the resource which means that investment potential is consid-
ered as available resources of legal entities. Financial, material, technical and labor re-
sources are the components of investment potential. The majority of supporters of this
approach use value assessment of the mentioned types of resources for measuring in-
vestment potential. However, such costing method is quite rough, because the main
focus is only on financial resources reflecting their cost measurement.
V. Shchelkunow [5] and C. Schulz [6], who are the market approach supporters,
associate investment potential with the demand for goods and services. Demand can
specify the volumes of investments for production, but it is essential to take into account
resources of the enterprise.
K. Pokataeva [7] has considered the investment potential as probabilistic possibility
of accumulation of the appropriate volume of resources for further investment by com-
pany. Although the researchers consider probability, they do not use the tools of prob-
ability theory.
T. Makukh [8] and T. Luehrman [9] in their works suggest the ability to achieve a
certain return on the resources used (regardless of their type), or from the standpoint of
assessing the economic results of current and future economic activity, the ability to
generate investment income. This definition characterizes the resultative approach. Es-
timation of investment potential is very close to DCF-method (Discounted Cash Flow),
but they are not similar and can produce different results. That is why assessment
method due to this approach is considered to be understudied.
Complex approach for solving solving different types of information asymmetry
problems between SMEs and other market participants aimed at simplifying investment
and innovation processes in Ukraine for providing available useful information about
the possible ways to attract financial resources was proposed by Yu. Sybirianska [22].
Capacitive approach revealed the ability of the subject of research (territory, enter-
prise, subsystem, etc.) "to absorb capital", which depends on several objective and sub-
jective factors, and is described in papers such scientists as O. Shelest [10], O. Goralko
[11], A. Kostonichenko [12] and J. Daggers & A. Nicholls [13]. Researchers face the
problem of gathering data. Their approach has describable character and does not allow
to evaluate the investment potential.
Previous studies did not consider all aspects of investment potential estimation.
Moreover, these researches are based on old fashioned, weak methods of assessment
and they demonstrate poor usage of statistical tools or probability theory. The consid-
ered methods are limited in using for assessment of investment potential on macro
level, which can show the role and the place of SMEs investment activity. Research has
been changing dramatically: the core focus is big data. These days the improved avail-
ability of data allows many investors (including SMEs) to make their decisions using
instruments of machine learning (ML).
Nowadays, ML has changed the investment landscape, taking into account that ML
and artificial intelligence may unleash new insights, like using data analysis before in-
vesting significantly in the technology.
In this study, we propose to use ML in SMEs investment potential estimation be-
cause it reveals the market structure, the effect of macro factors on SMEs, shows dif-
ferent drivers and losses of the economy due to technology changes and allows to assess
business environment. For instance, making million transactions everyday ML can de-
fine where spending money is growing or prices are rising, effecting mostly consumers.
This analysis can help consumers, policymakers and business representatives and other
leaders make smarter decisions.
3 The Main Characteristics of SMEs Investment Potential
Tendencies in Ukraine
Criteria for enterprise determination according to its size in Ukraine were changed by
adding the new definition for microenterprises due to adoption of the Law on Develop-
ment and State Support of Small and Medium Entrepreneurship in Ukraine in 2012
which amended the Commercial Code of Ukraine.
The State Statistics Services has started to generate data about the main indicators
of business environment using these new criteria for enterprise definition since the
mentioned date of law adoption.
Along with legislative changes such current urgent problems existed in Ukraine as
conflict in the east of the country and long-term lack of key reforms, the economy stag-
nation and recession could be observed as the main constraints for creation and imple-
mentation of SME support policy.
Besides, the positive moments of business development within 2012-2016 include
implementation of the key measures for business registration simplification, extension
of e-government services, elimination of trade technical barriers, EU standards adop-
tion and so on. On the other side, a lack of access to finance and real long-term strategy
for SME could be considered as the debilitating force for business development.
Though the structure of business in Ukraine should be analyzed regarding such im-
portant indicators in dynamics as number of enterprises, their annual turnover and num-
ber of employees (fig. 1).
Analyzing fig. 1 it can be specified that in 2016 in total SMEs made up for more
than 99 % of the legal entities in Ukraine. Overall, it is clear that while the growth rate
of SME number fluctuated, the general trend was downward from 2,2 million entities
in 2010 to 1,8 million entities.
In 2016, the private sector constitute in Ukraine was represented by 95,6 % of mi-
croenterprises, 2,6 % of small and almost 1 % of medium enterprises.
The chart from fig. 1 provides the dynamics of number of people employed in
Ukraine within the period of 2010-2016, which decreased from 8,4 million people in
2010 to 6,5 million people in 2016. Negative growth rate of employed people for SMEs.
is observed throughout almost all analyzed period of time
The share of SMEs in employment remained similar at approximately 77 %, here-
with the share of medium enterprises for employed individuals was about 32 %. There-
fore the significant imbalance can be observed as the share of medium enterprises is
lower than 1 % of total legal entities in Ukraine.
Fig. 1 also shows the positive dynamics in turnover increasing in Ukraine within
2010-2016 with the growth rate about 87 % in 2016 compared to 2010. But the highest
percentage has been marked for large and medium sized enterprises, equaling at about
30 % each.
Besides, it should be empathized that the share of large enterprises is less than 1 %,
but the share of its annual turnover is 3 times higher than share of microenterprise turn-
over.
100%
10,3% 8,8% 9,8% 10,6% 10,9% 12,0% 12,1%
90%
32,9% 11,3% 29,8% 10,4% 28,6% 10,7% 30,0% 10,9% 33,8% 11,1% 34,7% 12,5%
80% 35,2% 11,8%
70%
13,9% 38,5% 14,4% 14,1%
60% 13,2% 39,4% 40,0% 38,8% 12,9% 38,9% 11,7% 39,3% 13,0%
39,9%
50% 95,9% 94,5% 94,4% 95,1% 96,3% 96,8% 96,5%
40% 32,2% 32,0% 31,4%
31,7% 31,4% 32,2% 32,7%
30%
20% 39,0% 42,3% 39,5% 39,6% 39,1% 37,0% 35,6%
10% 22,3% 24,1% 24,9% 24,5% 21,8% 20,9% 19,6%
3,1% 2,9% 2,4% 2,6%
4,2% 4,3% 3,8% 0,8% 0,8%
0% 1,0% 1,2% 1,3% 1,1% 0,9%
Number
Number
Number
Number
Number
Number
Number
Employment
Turnover
Employment
Turnover
Employment
Turnover
Employment
Turnover
Employment
Turnover
Employment
Turnover
Employment
Turnover
2010 2011 2012 2013 2014 2015 2016
large enterprises medium enterprises small enterprises microenterprises
Fig. 1. General trend of business development in Ukraine within 2010-2016, % (Source: compiled by authors on the basis of [14])
To assess business development, it is reasonable to determine the investment potential
of enterprises by determination of structure their equity and liabilities (fig. 2).
5000000
4500000
4000000
3500000
3000000
2500000
2000000
1500000
1000000
500000
0
large enterprises
large enterprises
large enterprises
large enterprises
medium enterprises
medium enterprises
medium enterprises
medium enterprises
small enterprises
small enterprises
small enterprises
small enterprises
microenterprises
microenterprises
microenterprises
microenterprises
2013 2014 2015 2016
Equity current liabilities long-term liabilities
Fig. 2. Investment potential according to size of enterprise, thousands of hryvnas (Source: com-
piled by authors on the basis of [14])
Significant growth of capital can be observed for large and medium enterprises in 2016
in comparison with 2013 level, the growth rate equals 94,5 % and 54,6 % for each
group respectively.
The investment potential of small and microenterprises is much less than for large
and medium ones and mainly represented by current and long-term liabilities, the share
of equity is too insignificant (about 4-5 %). In comparison with large and medium en-
terprises the level of equity sufficiency equals about 36,6 % and 10,2 % respectively.
Fig. 2 confirms that investment potential of SMEs is inefficient which proved by the
structure of its liabilities which mainly represented by debt and not bank loans. That is
why the level of provision by financial resources of SMEs is excessively low.
The next stage of analysis of SMEs investment potential tendencies is to determine
the level of investment activity and its directions according to size of enterprises (fig.
3).
Fig. 3 demonstrates that the lowest level of investments is inherent to small and mi-
croenterprises, the main areas of which are investment to machinery and equipment.
0
20000
40000
60000
80000
100000
120000
large enterprises
medium enterprises
2010
small enterprises
Investment in Software
microenterprises
large enterprises
Investment in existing buildings
medium enterprises
2011
small enterprises
Investment in Machinery and Equipment
microenterprises
large enterprises
medium enterprises
2012
small enterprises
microenterprises
large enterprises
medium enterprises
2013
small enterprises
microenterprises
large enterprises
Investment in land
medium enterprises
2014
small enterprises
microenterprises
large enterprises
medium enterprises
2015
Investment in Construction and Reconstruction
small enterprises
microenterprises
large enterprises
medium enterprises
2016
small enterprises
Fig. 3. Areas for investment according to size of enterprise, thousands of hryvnas (Source: compiled by authors on the basis of [14])
Investment in Patents, Licenses, Trade Marks and other Intellectual Rights
microenterprises
The amount of investment for large and medium enterprises is almost 3 times higher
than for small ones, herein the main spheres of investment are investment in construc-
tion and reconstruction and machinery and equipment, that prove that financial re-
sources are mainly directed to production.
Investment into software, patents, licenses and trademarks is not widespread for en-
terprises in Ukraine.
4 Doing Business Index as a Tool of Business Environment
Assessment
As known, Global competitiveness index (GCI) is global research and the ranking of
countries accompanying it in terms of economic competitiveness [21]. But due to the
fact that the SME investment potential depends on factors of business environment
measured by indicators of Annual Doing Business Reports developed by the World
Bank Group it is advisable to consider the methodology of Doing Business rank. The
methodology consists of 10 groups of factors with its indicators shown in table 1.
Table 1. The essence and factors of “Doing Business” rank methodology
Factor The essence Indicators
Starting a busi- Identification of bureaucratic, legal constraints Procedure (number)
ness and costs required for entrepreneurs aimed at Time (days)
starting new business. Cost (% of income per capita)
Paid-in min capital (% of in-
come)
Dealing with Assessment of procedures, time and financial Procedures (number)
construction resources, connected with construction pro- Time (days)
permits cesses, obtaining permits and licenses, manda-
tory instruction, connection to utilities. Cost (% of warehouse value)
Building quality control index
(0-15)
Getting elec- Assessment of procedures, time and financial Procedures (number)
tricity resources, connected with getting electricity. Time (days)
Cost (% of income per capita)
Reliability of supply and
transparency of tariff index
(0-8)
Registering Assessment of procedures, time and financial Procedures (number)
property resources, connected with registration of prop- Time (days)
erty rights. Cost (% of property value)
Quality of the land admin-
istration index (0-30)
Getting credit Assessment of credit bureau coverage of indi- Strength of legal rights index
vidual entrepreneurs and legal entities, and (0-12)
their collateral, which presupposes the estima- Depth of credit information
tion of factors which can simplify access to index (0-8)
loans. Credit registry coverage (% of
adults)
Credit bureau (% of adults)
Factor The essence Indicators
Protecting mi- Assessment of protection level against illegal Extent of conflict of interest
nority inves- management of stock companies. Indices regulation index (0-10)
tors equal the sum of points for positive answers to Extent of shareholder govern-
relevant questions, e.g. one consent=one point. ance index (0-10)
Paying taxes Assessment of taxes and mandatory contribu- Payments (number per year)
tions which should be paid by companies. The Time (hours per year)
quality of tax administration and the level of Total tax and contribution rate
tax burden are determined. (% of profit)
Postfiling index (0-100)
Trading across Assessment of costs, including time, financial, Time to export: border com-
borders which should be paid due to export or import pliance (hours)
of goods. 20—foot container is considered as Cost to export: border compli-
typical situation. ance (USD)
Time to export: documentary
compliance (hours)
Cost to export: documentary
compliance (USD)
Time to import: border com-
pliance (hours)
Cost to import: border com-
pliance (USD)
Time to import: documentary
compliance (hours)
Cost to import: documentary
compliance (USD)
Enforcing con- Determination of the number of procedures, Time (days)
tracts term and costs of company required to debt Cost (% of claim)
collection from unscrupulous buyer-legal en- Quality of juridical processes
tity, which refused to pay for delivered goods, index (0-18)
citing its low quality in case when expertise
confirms the sufficient level of goods quality.
Resolving in- Determination of bureaucratic and legal con- Recovery rate (cents on dol-
solvency straints for an entrepreneur to overcome for lar)
company liquidation due to its bankruptcy and Time (years)
the main procedure and administrative bottle- Cost (% of estate)
necks of bankruptcy procedure. Assessment of Outcome (0 as piecemeal sale
set of company actions (terms, cost, the level and 1 as going concern)
of loan return) within bankruptcy procedure. Strength of insolvency frame-
work index (0-16)
Source: [15]
The rate is calculated on the basis of official statistical data and questionnaires of
companies, requirements for which are described in table 2. The mentioned rank repre-
sents the integrated indicator, which consists of 10 sub-indicators in different catego-
ries, which are important for entrepreneur activity. The meaning of rank which is the
closest to “top” position (1st rank in the list) shows better conditions for doing business
than ranks close to 190 in the list [16].
The typical company for Doing Business assessment is Limited Liability Company
located in the largest business center of country and 100 % domestically owned (more
detailed analysis is given in table 2).
Table 2. Requirements to companies according to indicator of “Doing Business” rate
Indicator/ Starting a Dealing with Registering Getting Paying taxes Trading Resolving insol-
Requirements business construction property credit across bor- vency
permits ders
Type of company – Limited Li- + + + + + + +
ability Company
City – the largest business center + + + + + + +
of country
Company 100 % domestically + + + + + + +
owned
Start-up capital equals 10 times in- – – - 102 times in- – -
come per come per
capita capita
Company activity does not in- + – – – –– –
clude foreign trade
Export volume of company – – – – – 10 % from –
annual turno-
ver
Company has real estate + – – – – – –
Company has building – – – – – – +
Company has land plot – + – – + – –
Company staff 10–50 em- 60 employees 50 employees Up to 50 60 employees - 201 employees &
ployees employees 50 suppliers
Annual turnover of company is 100 times – – – 1,050 times - -
not less income per income per
capita capita
Source: [15]
Table 3. Ukraine in “Doing Business” reports within 2006-2016
Doing Business 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Position in rate 124 128▼ 139▼ 145▼ 142▲ 145▼ 152▼ 137▲ 112▲ 87▲ 83▲
Starting a business – 101 109▼ 128▼ 134▼ 118▲ 112▲ 50▲ 47▲ 70▼ 30▲
Dealing with construction – 107 174▼ 179▼ 181▼ 179▲ 180▼ 183▼ 41▲ 139▼ 140▼
permits
Getting electricity – – – – – – 169 166▲ 172▼ 138▲ 137▲
Hiring – 107 102▲ 100▲ 83▲ – – – – – –
Registering property – 133 138▼ 140▼ 141▼ 164▼ 166▼ 149▲ 97 ▲ 64 ▲ 61 ▲
Getting credit – 65 68 ▼ 28 ▲ 30 ▼ 32 ▼ 24 ▲ 23 ▲ 13 ▲ 17 ▼ 19 ▼
Protecting minority inves- – 142 141▲ 142▼ 109▲ 109▬ 111▼ 117▼ 128▼ 87▲ 88▼
tors
Paying taxes – 174 177▼ 180▼ 181▼ 181▬ 181▬ 165▲ 164▲ 106▲ 107▼
Trading across borders – 106 120▼ 131▼ 139▼ 139▬ 140▼ 145▼ 148▼ 109▲ 109▬
Enforcing contracts – 26 46 ▼ 49 ▼ 43 ▲ 43 ▬ 44 ▼ 42 ▲ 45 ▼ 98▼ 98 ▬
Resolving insolvency – 139 140▼ 143▼ 145▼ 150▼ 156▼ 157▼ 162▼ 141▲ 141▬
Source: [17]
The rates for Ukraine in “Doing Business” reports are presented in table 3, from which it can be seen the upward trend for Ukraine
from 124 position in 2006 to 83 in 2016 demonstrates the positive dynamic reducing strong discontinuity. The positive changes are mainly
connected with such indicators as “Starting a business”, “Registering property” and “Paying taxes”, but other indicators prove the set of
complicated procedures for doing business in Ukraine.
The affirmative modifications relate to simplification of procedures of starting business, registering property and paying taxes exem-
plified as implementation of e-government services. All factors from table 1 and 3 have great impact for investment climate formation
that is why they should be considered for designing the model for forecasting the investment potential of SMEs.
Advantages of Doing Business methodology like available big databases, different
categories of parameters, possibility to overview risks and market potential, oppor-
tunity to make clear comparisons confirm the necessity to use data from Doing Busi-
ness rank for designing the mentioned model.
5 Machine Learning Model of SMEs Investment Potential
Estimating
The use of investment potential of SMEs is represented by allocating resources (liabil-
ities) to assets [18]. Obviously, investment potential is formed by equity and liabilities,
and its use by assets. Therefore, in the future, it is possible to use indicators of total
volumes of equity and liabilities for analysis. For determining and analysis of the im-
pact of business environment factors on SMEs it is advised to use such input data as
investment potential of SMEs in different European countries (Slovenia, Czech Repub-
lic, Estonia, Slovakia, Hungary, Latvia, Poland and Ukraine) and their indicators of
Doing business, mentioned above. We propose to use predictive models based on dif-
ferent methods: linear regression model and automatic relevance determination regres-
sion model.
The linear regression model is a type of modeling the ratio between the scalar y and
the vector variable x. Like other regression analysis methods, linear regression repre-
sents the probability distribution of y depending on x rather than the distribution of the
common probability y and x, which relates to the field of multivariate analysis.
In general, the linear regression model (one of the algorithms of ML) is defined as
follows:
y 0 1 x1 ... k xk u, (1)
where y is a dependent explanatory variable, (x1, x2, ..., xk) is an independent explan-
atory variable, u is a random error, the distribution of which in the general case depends
on independent variables, but whose mathematical expectation is zero [18]. The de-
pendent variable in our case is the value of the investment potential of the SMEs. Inde-
pendent explanatory variables in this paper are indices of business environment factors:
x1 – starting a business, x2 - dealing with construction permits, x3 - registering prop-
erty, x4 - getting credit, x5 – protecting minority investors, x6 – paying taxes, x7 -
trading across borders, x8 - enforcing contracts, x9 - resolving insolvency,
According to this model, the mathematical expectation of a dependent variable is a
linear function of independent variables:
( y) 0 1 x1 ... k xk u. (2)
The vector of parameters
0 , 1 , ... k is unknown and the problem of linear
regression is to evaluate these parameters based on some experimental values
yi (x1, x2,…xk). For some n experiments, there are known values {𝑦𝑖 , 𝑥𝑖1 , … , 𝑥𝑖𝑝 } 𝑛𝑖= 1
of independent variables and the corresponding value of the dependent variable. Ac-
cording to the model definition for each experimental case, the dependence between
the variables is determined by the formulae:
𝑦𝑖 = 𝛽0 + 𝛽1 𝑥1,𝑖 +. . . +𝛽𝑘 𝑥𝑘,𝑖 + 𝑢𝑖 ,
or in matrix notation:𝑦 = 𝑥𝛽 + 𝑢,
where
(3)
On the basis of these data, the value of the parameters (𝛽0 , 𝛽1 , … , 𝛽𝑘 ), is to be estimated
as well as the distribution of a random variable [18]. These models are chosen because
of their capacity to give the result of the analysis with the slightest error. For a specific
group of enterprises (small or medium sized), the experimental way was to determine
its model for predicting data.
These models are also implemented in the library of machine learning SCIKIT-learn.
In addition, this library contains methods for evaluating the obtained results, which are
used in the analysis of the impact of business environment factors on the development
of the SMEs. These include: Mean absolute error, Mean squared error, and R² (R² score,
the coefficient of determination).
For accurate analysis, it is necessary to normalize the entire amount of data that in-
clude equity and liabilities. Normalization is the process of analyzing ratios in order to
identify and eliminate abnormalities of modification. These anomalies can be elimi-
nated by splitting the initial relation into two or more new relations. The elimination of
anomalies is carried out according to the following formula:
Y Ymin
Ynorm , (4)
Ymax Ymin
where Ynorm is the normalized value, Y is the actual value, Ymin is the minimum
value for the total volume of data, Ymax is the maximum value from the total amount
of data.
By choosing models to analyze the investment potential of different SMEs, each of
them (model) needs to be trained for a variety of data: first by training (to generate
coefficients of variables), and then - testing. Whereas, data processing is time-consum-
ing and is to be optimized. The algorithm not only accelerated the process of data pro-
cessing, but could also make such a methodology suitable for analyzing the impact of
business environment factors on SMEs. The methodology was formed by the Python
programming language.
Thus, the developed models made it possible to determine the influence of factors
of the business environment on the SMEs investment potential. This technique allows
to assess the factors which increase their investment potential, and which reduce it. The
evaluation results may form the recommendations basis for improving the business en-
vironment that affects the activities of its actors, including SMEs.
In general, this allows to consider the factors which affect the development of SMEs,
outlining the main areas of creation, support and improvement of factors of the business
environment that determine the activities of SMEs.
This shows the best results of the forecast (the lowest error (app. 0.02) and the high-
est communication factor (R2=0.72) between the test and forecast figures). Data, ob-
tained from official sites of the Organization for Economic Development and Cooper-
ation, national regulators of the countries, used, have been normalized by the formula
(4). Analysis of the impact of the 9 business environment factors on SME investment
potential was made by using the SCIKIT-learn computer library and the Python pro-
gramming language [19, 20].
Thus, the influence of factors of the business environment on the formation and use
of investment potential of small businesses best describes the model of determination
of regression with automatic determination of relevancy for small sized enterprises,
while classical linear regression - investment potential of medium businesses.
In the learning process the prediction models and the coefficients of the equation
were obtained, the absolute value of the module which characterizes the degree of in-
fluence on the investment potential, and their sign indicates the nature of the effect ("+"
- increases the potential, "-" - reduces the potential).
The results of the modelling (table 4) show that such factors as starting business,
protecting minority investors, paying taxes, enforcing contracts (because their coeffi-
cients have positive sign) increase investment potential both of small and medium busi-
nesses. The value of coefficient demonstrates the impact strength. The most crucial
factor in forming the investment potential of small business is starting business
(6.09e+05), while for medium businesses investment potential is paying taxes
(6.3e+08). Paying taxes also plays a significant role due to the value of its indicator
(4.98e+00) for small enterprises, starting business has almost the same value for me-
dium sized enterprises (4.06e+01). Enforcing contracts also positively influences both
small and medium businesses and their values are close to each other (2.14e+00 and
2.64e+00). The least positive impact on formation of SMEs investment potential is
made by enforcing contracts factor.
Conversely, there is set of factors which reduce the investment potential of SMEs
(in the models they have negative sign). Resolving insolvency (-9.94e-07 and -
9.70e+00) and getting credit (-8.19e+03 and -8.44 e +04) are among the most destruc-
tive factors which sway on formation of SMEs investment potential. This fact proves
poor performance with low level of banking SMEs loaning. One more common influ-
ence but in different degree (for small businesses - (-1.81e-03) and for medium enter-
prises - (-5.91e+00)) is trading across the borders. Table 4 shows that for medium sized
businesses export opportunities are result forming. In the case of export strategy ab-
sence, the activity of medium enterprises remains in the lowest. Inability to export in-
hibits the competitiveness of medium enterprises of Ukraine in the global business en-
vironment.
Although most factors are common for all types of SMEs, at the same time models
help to observe some dissimilarities.
Table 4. Specification of models for forecasting the investment potential of small and medium
enterprises*
The degree of influence The degree of influence
of business environ- of business environ-
ment factors on small ment factors on me-
enterprises investment dium-sized enterprises
Designation of factors in the program code
Indicators of business envi- potential (automatic investment potential
ronment factors relevance determina- (classical linear regres-
Designation of factors in the formula
tion regression model): sion):
YSE=(6.09e+05)*X1+
(3.05e- YME= (4.06e+01)
01)*X2+(4.0.e+00)*X3+ *Х1 +
(-8.19e+03)*X4+ (-7.48e+00)*Х2 +
(0.25e- (-1.39e-+01)*Х3+(-
03)*X5+(4.98e+00)*X6 8.44 e +04)*Х4 +
+ (2.44e-01)*Х5+
(-1.81e-03)*X7+ (6.3e+08)*Х6+
(2.14e+00)*X8+ (-5.91e+00)*Х7+
(-9.94e-07)*X9 (1) (2.64e+00)*Х8+
(-9.70e+00)*Х9 (2)
Increasing Decreasing Increasing Decreasing
(0;+∞) (-∞;0) (0;+∞) (-∞;0)
Starting a business Х1 S_B 6.09e+05 … 4.06e+01
Dealing with construction Х2 D_C 3.05e-01 … … -
permits 7.48e+00
Registering property Х3 R_P 4.07.e+00 … … -1.39e-+01
Getting credit Х4 G_K … -8.19e+03 … -8.44 e +04
Protecting minority investors Х5 P_MI 0.25e-03 … 2.44e-01 …
Paying taxes Х6 P_T 4.98e+00 … 6.3e+08 …
Trading across borders Х7 T_B … -1.81e-03 -5.91e+00
Enforcing contracts Х8 E_C 2.14e+00 … 2.64e+00 …
Resolving insolvency Х9 R_I … -9.94e-07 … -
9.70e+00
*Designed by authors
Thus, dealing with construction permits benefits investment potential of small en-
terprises (3.05e-01), whereas this factor has negative influence on the same indicator
of medium businesses (-7.48e+00). The next distinguishing factor is registering prop-
erty: in the case of small enterprises the result of registering property has positive in-
fluence on investment potential (4.07.e+00), while, medium businesses experience the
negative impact (-1.39e-+01).
Therefore, the influence of business environment can differ for small and medium
enterprises, confirmed by the fact that state and strategy policy for small and medium
business should be different.
6 Conclusions
Investment potential of SMEs can be estimated with the help of ML tools, exemplified
as developed models for small and medium-sized enterprises assessment in particular.
These models allow to determine which factors of business environment have impact
on investment potential of SMEs. The decisive feature of these models is not only to
forecast investment potential but also to measure the degree of influence of each con-
sidered factor.
The results of the assessment can be used by policymakers and public authorities
paying attention to policy directions constraining business development of SMEs. Get-
ting credits should be supported by government in different ways (by monetary policy
with decreasing interest rate, or its return to the SMEs; by tax policy with incentives
for those SMEs which have credit pressure; developing non-banking funding etc.). Ac-
cording to the results resolving insolvency also reduces the investment potential of both
small and medium-sized enterprises. The required changes in legislation should protect
both creditors as well as bankruptcy enterprises from raiders.
The mentioned models can also help to identify the strengths and weaknesses of
SMEs activity. The calculated investment potential should be used in processes of strat-
egy formation, which determines the axis and performance of SMEs. Foreign investors
can consider these results in decision making for investment financial resources into
some economy or abstaining from it.
Statistical authorities also have to measure the indicators of the SMEs investment
potential and its elements and Doing business indicators more often (with 4-time per
year frequency). This will allow to make more precise forecasting and smarter decision-
making.
These prediction models can be used for evaluation of investment potential of SMEs
not only in Ukraine, but also in countries like Slovenia, Czech Republic, Estonia, Slo-
vakia, Hungary, Latvia, Poland, because they have some similar conditions for running
business.
References
1. Leonov, S.: Investment potential of the banking system of Ukraine. Sumy. DBA "UABS
NBU" (2009).
2. Eklund, J.: Theories of Investment: A Theoretical Review with Empirical Applications
(2013), http://entreprenorskapsforum.se/wp-content/uploads/2013/03/WP_22.pdf, last ac-
cessed 2018/01/03.
3. Yaremchuk, N.: Investment potential: theoretical foundations of formation. Digest of the
Lviv State Financial Academy 18, 225–231 (2010).
4. Dieterlen, P.: Limites de l'investissement global et potentiel d'investissement. Revue
économique. Volume 6, (3), 448-475 (1955).
5. Shchelkunow, V. et al.: Marketing of the regions: investment aspects. Kyiv. Naukova dumka
(2005).
6. Schulz, C.: Investing in the Global Marketplace: Opportunities in International Equities
(2014) https://content.pncmc.com/live/pnc/corporate/pncideas/articles/WhyInternational-
Final.pdf, last accessed 2018/01/05.
7. Pokataeva, K.: Theoretical basis for defining the categories of "investment potential" and
"investment attractiveness". Utilities of cities 75, 264-270 (2007).
8. Makukh, T.: Investment potential as a factor in increasing the efficiency of using financial
resources. Bulletin of the National University of Water Management and Nature Manage-
ment 1 (41), 175-186 (2008).
9. Luehrman, T.: Investment opportunities as real options: Getting start on the numbers. Har-
vard business review, July-August. Reprint. 98404 (1998), https://hbr.org/1998/07/invest-
ment-opportunities-as-real-options-getting-started-on-the-numbers, last accessed
2018/01/08.
10. Shelest, O.: Theoretical aspects of formation of structure of investment potential of air-
transport enterprises. Economics and Management: Collection of scientific works. Lutsk
National Technical University 7 (26) Part 3, 351-357 (2010).
11. Goralko, O.: Influence of the Investment Potential of the Ukrainian Banking System on the
Stabilization and Development of the National Economy. Bulletin of Lviv Commercial
Academy 37, 91-94 (2011).
12. Kostonichenko, A.: Management of investment adaptability of construction companies
(Doctoral Thesis) (2003), http://mgsu.ru/, last accessed 2018/01/10.
13. Daggers, J., Nicholls, A.: The Landscape of Social Impact Investment Research: Trends and
Opportunities (2016). https://www.sbs.ox.ac.uk/sites/default/files/research-pro-
jects/CRESSI/docs/the-landscape-of-social-impact-investment-research.pdf, last accessed
2018/01/08.
14. Reports of business activity of enterprises in Ukraine, 2010-2016. Official Statistics Service,
http://www.ukrstat.gov.ua, last accessed 2018/01/03.
15. Doing Business. Measuring business regulations. Methodology, http://www.doingbusi-
ness.org/methodology, last accessed 2018/01/12.
16. Ivashchenko, A., Orlova, N.: Comparative analysis of some EU member states and EU as-
sociated countries to identify the phenomenon of business development in post-socialist
countries. Economic annals XXI 163 (1-2 (1)), (2017). DOI:
https://doi.org/10.21003/ea.V163-04.
17. Doing Business. Measuring business regulations. Rankings., http://www.doingbusi-
ness.org/rankings, last accessed 2018/01/10.
18. Polishchuk, Y.: Non-banking financial institutions on investment market. Chernihiv, Cher-
nihiv National University of technology (2016)
19. Muller, A.C., Guido, S.: Introduction to Machine Learning with Python. Moscow, Sankt
Petersburg. Kyiv. Dialectica (2017).
20. Coelho, L.P., Richert, W.: Building Machine Learning Systems with Python. Moscow.
DMK Press (2016).
21. Sardak, S., Samoylenko, A.: National Economies Intellectualization Evaluating in the World
Economy. Economic Annals-XXI 9-10(2), 4-7 (2014).
22. Ivashchenko A.I., Sybirianska, Yu.V., Polischuk Ye.A: Information and Communication
Platform as a Complex Approach for Solving Information Asymmetry Problems. CEUR-
WS Conference Proceedings, pp. 111-127. (2017),
http://icteri.org/sites/default/files/ICTERI-2017-Proc-Vol-CEUR-WS-1844.pdf, last
accessed 2018/01/09.