• About Us +
• Editorial Board +
• For Contributors +
• Journal Search +
Journal Search Engine
ISSN : 1598-7248 (Print)
ISSN : 2234-6473 (Online)
Industrial Engineering & Management Systems Vol.18 No.3 pp.454-462
DOI : https://doi.org/10.7232/iems.2019.18.3.454

Selecting Service Providers with Neural Networks: Evidence from Car Service Providers

S. M. Khairova, V. A. Kovalev, B. G. Khairov*, A. V. Shimohin
Financial University under the Government of the Russian Federation, Siberian State Automobile and Highway University (SibADI), Russia
Financial University under the Government of the Russian Federation, Russia
Financial University under the Government of the Russian Federation, Laboratory of Economic Research of the Omsk Region of the IEEPP SB RAS, Russia
Omsk State Agrarian University, Omsk, Russia
Corresponding Author, E-mail: bari.khairov@yandex.ru
May 29, 2019 June 16, 2019 June 24, 2019

ABSTRACT

The research goal is to test the ability of neural networks to solve management problems, e.g. to select a car service provider. To achieve this, the authors proposed a reliability scale of a car service provider and obtained a dependence to assess this reliability. These procedures provided the data that were used to train a neural network. The authors applied the expert method to obtain the weights of the parameters describing car service providers. The opinions of experts were used to compile the characteristics of the car service providers. These data helped train neural networks to identify the most optimal parameters for solving this problem – to select of a car service provider. The conducted research confirms the ability of neural networks to solve management problems of choosing service providers. The study revealed the most appropriate type and structure of the neural network for dealing with such a management task. The findings of the study can be used to create a software product that can help managers choose a service provider. Besides the authors proposed the criteria for identifying changes in the quality of services provided.

1. INTRODUCTION

The main functions that transport should perform are timely, efficient and full coverage of the transport needs of the national economy and people, as well as increasing its economic efficiency in the context of import substitution. The flawless operation of transport and the quality of its services depend on the technical condition of transport, which in turn is ensured by service providers. Thus, service providers play a significant role.

A car service provider is usually selected according to experts’ opinions. However, not all logistics infrastruc-ture enterprises can use the knowledge and experience of the experts capable of making a good choice of a supplier (Khairov and Karpova, 2016;Khairov and Blinov, 2015;Sharif and Butt, 2017;Otter et al., 2017;Tyapukhin, 2013). In addition, in large cities there is a huge selection of suppliers, which makes it hard to find a reliable one.

Neural networks can be used to solve various economic tasks as they can learn on training data sets containing the knowledge and experience of experts (Mahmood et al., 2019;Salih, 2019). These data may be useful for various enterprises of the logistics infrastructure. The mathematical apparatus – neural networks – is developed and applied for solving various problems of the national economy. Currently, many authors (Falat and Pancikova, 2015;Hadley, 1964;Farzadnia et al., 2017;Ahmadi et al., 2018) are exploring ways of using the neural network apparatus in the economy, for solving such tasks as forecasting de-mand, prices, making decisions about granting a loan (the task of classifying a client) and others. The advantage of neural network modeling stems from their ability to learn, which enables to obtain accurate results with any number of factors and their values. The learning process and the operation principle are discussed in detail in the author’s article (Khairova et al., 2015;Muhammad, 2018;Gumel, 2017). Thus, the relevance of developing autonomous neural networks for selecting a service provider is obvious.

The research goal is to test the ability of neural net-works to solve management problems, e.g. to select a car service provider. To achieve this, the authors proposed the reliability scale of a car service provider and obtained a dependence to assess this reliability.

These procedures provided the data that were used to train a neural network. Then, the authors trained and tested the RBF neural network. It gave the most accurate result that was closest to the expert opinions, compared to MLP neural networks. The authors determined the architecture of the RBF neural network and obtained its weights, as well as formulated the criteria that can be used to assess the quality of a service provider and, if necessary, to apply them to protect the interests of the enterprise in the event of claims to the service provider.

In the future, the study findings can be used to develop a software product based on the obtained neural network. This will allow logistics infrastructure enterpris-es to apply experts’ knowledge when choosing a car service provider, to classify them accurately and to make a decision on choosing one. The proposed service quality factors will help protect the interests of the enterprise.

2. LITERATURE REVIEW

The theory and practice of outsourcing have been studied by (Dobronravov, 2003). James Brian Quinn considered the outsourcing at Indian manufacturing companies and its impact on their productivity (Quinn, 2000). Also, in (Fixler and Siegel, 1999) analyzed the influence of outsourcing on productivity.

Increasing global competition in the markets of goods, services and capital entails a change in the national and global freight traffic, stricter requirements for the quality of transport services. In the context of the global economic crisis, leading, among other things, to a reduction in the transport services market, almost all transport companies are taking greater efforts to attract clients, applying various methods of price and non-price competition to achieve this goal. In these difficult conditions, more transport companies are paying attention to the reasonable improvement of the quality of services (Zitkiene and Dude, 2018;Rustamova, 2013). This trend should reflect the require-ments of the present day, and therefore, allow increasing their competitiveness (Novikova et al., 2016;Mazur et al., 2016). It should be noted that Western and Russian researchers primarily focused on the im-provement of product quality, whereas the quality of services (including transport) was and is a much less investigated topic. However, the choice of a service provider is a problem that has not been considered yet.

3. METHOD

Neural network modeling represents the methodo-logical base of the research. The study involved the methods of collecting primary and secondary information, statistical data analysis, and the expert method.

The expert method was used to obtain the weights of the parameters characterizing car service providers. The heads and engineers of enterprises possessing and actively using motor vehicles acted as the experts. After the decision on the outsourcing of car services is made, one should evaluate potential providers of these services (Cabral et al., 2014; Kalendzhyan, 2003; Christopher, 2005;Dybskaya et al., 2008). In addition to the cost and the range of services offered by the supplier, the authors also suggest considering such parameters as: work experience in this market, business reputation, etc. The proposed parameters are presented in Table 1. The authors conducted a survey among the experts (the directors and engineers of enterprises using car services as the transport of the enterprises) and obtained weights of these criteria. The questions of the survey are given in Table 1. The authors propose the following scale for evaluating service providers (see Table 2).

The parameter describing the reviews was calculated as the difference between the sums of positive and negative reviews (Formula 1):

$R 3 = ∑ N p − ∑ N o$
(1)

where R3 is the value of the parameter describing the number of reviews about the service provider;

• Np is a positive review;

• No is a negative review.

The coefficient of the consistency of expert answers was calculated using STATISTICA program (Figure 1).

In recent years, neural network technologies have been used in a number of ways. In general, the operation of neural networks can be described as follows: input signals (different source data) are received through multiple input channels (Figure 1). The signal passes through the connection (synapse), which has a certain weight. Each neuron has a certain threshold value. Then a weighted sum of inputs is calculated, from which the threshold value is subtracted, and the result represents the activation value of the neuron. Using the activation function, this value is converted and the output signal is obtained. Figure 2

Neural networks are not programmed in the usual sense of the word, they are trained. The ability to learn is one of the main advantages of neural networks over traditional algorithms. Technically, learning implies finding the connections coefficients between neurons. During the learning process, the neural network is able to identify complex dependencies between input and output data, as well as to perform generalizations. This means that if the training is successful, the network will be able to obtain the correct result based on data that were not available in the training sample, as well as incomplete and/or "noisy", partially distorted data.

Let us enter the data (Table 3) into Statistica program. Column 5 is added for the training and control sample. Next, let us start the Neural Network, the tool for developing neural networks in Statistica program, and select the type of task – classification.

Next, let us distribute the data as follows: the characteristic of the service provider is the loss function; lifespan, large clients and reviews are continuous inputs.

These data were used to train the neural networks with Statistica 13 software. This allowed identifying the optimal parameters of the neural network for choosing a car service provider.

The input data for the neural network are the parameters proposed in the study, i.e. the ones characterizing the service provider: the lifespan in the market, large enterprises as clients, and available reviews. The scale of reliability was proposed to characterize the service providers. This characteristic includes the listed parameters and the dependence for assessing reliability. As a result, the neural network will give one of the proposed characteristics to the supplier: an absolutely reliable service provider, a reliable service provider, a fairly reliable service provider, a fairly unreliable service provider, an unreliable service provider, an absolutely unreliable supplier. The more reliable ones should be the first to choose, with an absolutely reliable provider on top.

After training, the neural network could classify service providers according to the proposed reliability scale, and the result complied with the experts’ opinion. The architecture of the neural network was determined with least error, as well as its parameters (weights, activation function). Also, its operation was tested on various input data.

The ability of neural networks to learn and solve various problems related to classification and forecasting determines its application in economics. According to the hypothesis of the study, a neural network trained in a training sample that is based on the experts’ assessment of service providers can accurately classify service providers to meet the needs of logistics infrastructure enterprises.

The application of neural network modeling in economics is expanding. The ability of neural networks to learn allows them to use the knowledge and experience of experts, which determines high-quality results of the neural networks operation.

The conducted expert method and the created scale of preferences of service providers were used for neural network modeling. This provided various versions of architectures, weights and activation functions of neural networks. However, the performance analysis showed that this problem is best solved by a RBF neural network, whose operation was tested in Statistica 13 program.

When choosing a service provider, it is necessary to take into account all outsourcing costs, including those associated with the possible transportation of the equip-ment to the repair site (Kushnirov, 2005;Shukla, 2017;Kvet and Matiasko, 2018).

Costs for car service outsourcing can be expressed by Formula 2:

$C i = ∑ j k S i j × N i j + S D i + Z i$
(2)

where

• Ci is the costs of using the services of the i-provider;

• k is the number of types of services provided to the company;

• j = {1…k};

• Sj is the cost of the j-th type of services of the i-th provider;

• N is the number of the required services of the j-type offered by the i-th provider;

• SD is the cost of transporting the car to the repair site of the i-th provider, if this service is offered by the company;

• Z is the cost of spare parts of the i-th provider.

Thus, when choosing the i-th provider, it is necessary to consider the conditions under which the number of required services (Ni) reaches the maximum value, and the costs of using the i-th service provider – the minimum value.

The results obtained comply with the model (3). Thus, the RBF neural network was trained in Statistica program using the experts’ data.

When concluding a contract for the provision of services, one should develop a mechanism for assessing the work quality. Car services can be assessed using the following parameters:

$K 1 = t n + 1 / t n$
(4)

where K1 is the waiting period;

• tn+1 is the waiting time: the time spent on preparing documents for the repair and reception of the car for the (n+1)-th period of using the services of the provider;

• tn is the same time, for the n-th period of using the services of the provider.

$K 2 = t n + 1 / t n$
(5)

where K2 is the time of rendering the service;

• tn+1 is the time of rendering the service for the (n+1)-th period, tn is the time of rendering the service for the nth period.

$K 3 = m n + 1 / m n$
(6)

K3 is the number of failures;

• mn+1 is the number of failures for the (n+1)-th period, mn is the number of failures for the n-th period.

If in the previous period the company did not use the services of the provider and repaired the cars by itself, then the parameter K1 is not calculated.

When these parameters decrease, one can talk about the improvement of the service quality. The process of providing car services can be considered successful with the following values of the parameters:

$K 1 < = 1 , K 2 < = 1 , K 3 < = 1$
(7)

These parameters can be used to justify changes in the contract for the provision of services regarding the cost or other aspects.

4. RESULTS

A service provider is selected according to the opinion of the experts (Kushnirov, 2004; Shestoperov, 2014). To ensure the consistency in the provider’s selection, the authors proposed a number of parameters, and a scale of preferences was developed. Having applied these methods, the authors obtained weights, and the consistency of the expert opinion was verified by the Kendall coefficient of concordance which estimated 0.73. This indicates that the expert opinions were consistent.

A fragment of the result of the expert survey is given in Table 4.

Thus, car service providers can be assessed by a dependency that determines the most suitable service provider.

$N i = 0.37 × R 1 + 0.23 × R 2 + 0.4 × R 3$
(3)

where Ni is the reliability of the i-th service provider;

• R1 is the parameter describing the lifespan of the service provider in the market;

• R2 is the parameter describing the number of large clients the service provider has;

• R3 is the parameter describing the number of reviews on the service provider.

This demonstrates the possibility of using neural networks for selecting a service provider. To train a neural network for such a task, the authors propose the following algorithm:

Let us create a scale of preferences for car service providers (Table 5).

The second step is to model the characteristics of various service providers. The parameters were generated randomly by the obtained model (3) and the service provider’s characteristic was selected according to Table 5 and the value of Ni. The results are presented in Table 6.

To solve this problem, a RBF neural network was built, and the operation of this neural network was tested in Statistica program.

Logistics and tangential activation functions are commonly used for classification problems in neural networks. Next, let us select the required parameters: the type of neural network (MLP – Multilayer perceptron), the number of hidden neurons, and training epochs.

Finally, the authors found out that the MLP neural network showed a low result (Table 7), with the performance factor of 90%, that is, in 90 cases out of 100 the model gives the correct result. The quality of the forecasting was improved by using a different type of neural network architecture.

Next, let us select the RBF type (radial basis function network) to solve this task.

More details are presented in Table 8.

It is worth noting that the fourth of the obtained models of the RBF type produced a 100% result. The activation function of this network is a Gaussian function (Ahmadi et al., 2018).

The parameter sensitivity analysis is presented in Table 9. It is not necessary to use the parameter “large clients” in further study

Also, Statistica program allows one to transfer weights (synapses) of the neural network to other editors, which makes it possible to develop and use an autonomous neural network for further research. The work of the neural network was tested (Table 10).

5. DISCUSSION

In this study the authors aimed to test the ability of neural networks to solve management problems, e.g. to select a car service provider. To achieve this goal, the authors determined the preference criteria for car service providers. The study proposes the parameters for evaluating service providers: lifespan on the market, large enterprises as clients and the availability of reviews on the company. This set of parameters is debatable. The authors believe that it is sufficient for solving this problem. However, the issue requires further research to obtain the most objective parameters. The set of parameters received no criticism from the experts (at the further stage of research, collecting experts opinion).

By applying the expert method the authors could build a model (3) which can be used to evaluate the service provider. This included determining the reliability of the company. In addition, the sensitivity analysis showed that the parameter “large enterprises as clients” has the least impact on the reliability evaluation of the service provider.

Next, the authors proposed a scale of preferences for assessing service providers (Table 4). This makes it possible to choose from the most reliable service provider from the set, that is, it can be used as a selection tool. To train the neural network, the authors created a training sample with a random set of parameters of service providers and a reliability value in accordance with model 3 which reflected the results of the expert method.

Neural networks analyzed this sample (Table 5) to assess the reliability of suppliers by the proposed parameters. Two types of MLP neural networks with a tangential activation function and a two-layer RBF perceptron with a Gaussian activation function were studied. Having evaluated the performance of neural networks, the authors determined that the RBF neural network with a 3-20-7 structure was best at performing this task. It gave the result which in 100% complied with the model (3), that is, it was the same as the opinions of the experts.

This research result can be used to develop a software product based on this neural network. It may help managers select a car service provider, that is, to provide support in decision-making. It should be noted that this result of neural network modeling has a drawback – a small training sample. However, the research goal was to test the ability of neural networks to assess service providers.

Different authors have proposed various methods for assessing the effectiveness of outsourcing (Johnson et al., 2002;Lukinsky et al., 2012;Zhatkin, 2018). Some researchers (Ahmadi et al., 2014) consider one-criterion approaches to assess economic efficiency. Papers (Prokofieva and Khairov, 2016; Prokofieva, 2010; Lambert and Burduroglu, 2000;Bentley and Bossé, 2018) propose multi-criteria approaches to evaluating the effectiveness of outsourcing and discuss their drawbacks. The latter may be due to the fact that the lower results for some criteria can be compensated by higher values of others. For instance, the final value may indicate a positive effect, although the values of the most important cri-teria will be worse. Thus, it is necessary to conduct fur-ther research on ways of evaluating the efficiency of outsourcing services, which should include a mechanism for determining economic efficiency and parameters reflecting the quality of the services provided.

The authors have proposed a set of parameters for assessing the operation of a car service provider which will enable to identify the potential changes in the quality of services regarding the most important criteria: waiting time, service time, and the number of failures. The result of this assessment can be applied in case of disputes between the company and the service provider. These criteria are also debatable and require further analysis and testing. However, there is obvious demand for such a mechanism in the developing market for services and outsourcing.

6. CONCLUSION

The conducted expert analysis allowed the authors to obtain weights of the parameters characterizing car service providers and to develop a scale for their evaluation, giving an opportunity to assess the service provider before starting to work with this company. As a result, a company obtains a tool for making an informed choice of a car service provider according to the following parameters: lifespan on the market, reviews and a number of large clients. The analysis showed that the parameter “large clients” is less significant in the selection procedure, while “reviews” is the most important. This is also confirmed by neural network modeling.

The article also proposes the criteria for work evaluation that reflect changes in the quality of the service: waiting time, time of service performance, the number of failures, which can be used to protect the interests of the company in the event of claims to the service provider.

The study tested the ability of neural networks to solve management problems, e.g. the selection of a ser-vice provider. The paper presents the data that were used to obtain the architecture, weights and other parameters of the neural network, which allowed solving this task, i.e., to determine which of the considered groups the supplier belongs to. Such an approach will make it possible to give priority to those service providers that have been classified as more reliable, with a focus on absolutely reliable ones.

The carried out research and the obtained results enabled the authors to conclude that neural network algorithms are an effective tool for an economist since the accuracy of neural network forecasting considerably exceeds the accuracy of the forecasts made by traditional methods. This conclusion was confirmed by the results of the classification of car service providers performed to select the best one with the neural network that had the two-layer perceptron RBF architecture and Gaussian activation function.

The results of the research can be used as the basis for the development of a software product using neural networks. This will allow logistics infrastructure compa-nies to use the knowledge of experts when selecting car service providers, to obtain qualitative results of their classification and to make decisions on the choice of a service provider.

ACKNOWLEDGMENT

The article presents the findings of the research carried out using the budget funds under the government assignment of the Financial University.

Figure

Estimation of the consistency coefficient and other parameters.

The model of a neural network.

Table

Questionnaire

The scale for evaluating car service providers

Data in statistica software

The result of the experts survey

The preference scale of car service providers

Characteristics of various service providers

Simulation results in the MLP network

Simulation results in the RBF network

Parameter sensitivity analysis

The output of the neural network

REFERENCES

1. Ahmadi, A. A. , Sirayi, A. B. , and Moghadasan M. H. (2014), Information technology; a facilitator for improving dynamic capabilities through knowledge management utilization, UCT Journal of Management and Accounting Studies, 2(2), 27-36.
2. Ahmadi, F. , Rahimi, M. , and Rezaei, A. (2018), Study of relation between business model and sensemaking decisions, Journal of Humanities Insights, 02(02), 99-108.
3. Bentley, B. and Bossé, M. J. (2018), College students’ understanding of fraction operations, International Electronic Journal of Mathematics Education, 13(3), 233-247.
4. Cabral, S. , Qelin, B. and Maia, W. (2014), Outsourcing failure and reintegration: The influence of contractual and external factors, Long Range Planning, 47(6), 365-378.
5. Christopher, M. (2005), Logistics and Supply Chain Management, Macmillan Publishers India, Ghaziabad.
6. Dobronravov, A. N. (2003), Subcontraction and outsourcing, Financial Management, 3, 43-48.
7. Dybskaya, V. , Zaitsev, E. , Sergeev, V. , and Sterligova, A. (2008), Logistics. Integration and Optimization of Logistics Business Processes in Supply Chains (ed.) Prof. V. I. Sergeyev, Moscow: Eksmo.
8. Falat, L. and Pancikova, L. (2015), Quantitative modeling in economics with advanced artificial neural networks, Procedia Economics and Finance, 34, 194-201.
9. Farzadnia, E. , Hosseini, Z. , and Riahi, A. (2017), Study of hospital quality management and improvement rates in the hospitals, Journal of Humanities Insights, 01(01), 7-11.
10. Fixler, D. J. and Siegel, D. (1999), Outsourcing and productivity growth in services, structural change and economic dynamics, Elsevier, 10(2), 177-194.
11. Gumel, F. (2017), The effects of European Negotiatory state of Turkey on local management, Journal of Humanities Insights, 1(2), 57-62.
12. Hadley, G. (1964), Nonlinear and Dynamic Programming, Addison-Wesley, Reading, MA.
13. Johnson, J. , Wood, D. F. , Wordlaw, D. L. , and Murphy Jr., P. R. (2002), Modern Logistics (7th ed.), Williams, Moscow.
14. Khairov, B. G. , and Blinov, A. O. (2015), Regulation of financial flows of technological modernization in the conditions of liberalization of the country’s economy, Bulletin of Siberian State Automobile and Highway University, 3(43), 99-103.
15. Khairov, B. G. and Karpova, S. V. (2016), Logistics for Bachelors, INFRA-M, Moscow.
16. Khairova, S. M. , Biryukov, V. V. , Romanenko, E. V. , and Khairov, B. G. (2015), Cyclic-temporal competitive advantages of the national economy and entrepreneurship development, Mediterranean Journal of Social Sciences, 6(4), 64-71.
17. Kushnirov, V. B. (2005), Aspects of taking a decision on outsourcing, Proceedings of the the 3rd Interregional Conference Entrepreneurship in industry: ways of development, Moscow, Available from: http://subcontract.ru/Conf2004/Disk/section3/section3_10.htm
18. Kvet, M. and Matiasko, K. (2018), Temporal data performance optimization using preprocessing layer, Journal of Information Systems Engineering & Management, 3(2), 13.
19. Lambert, D. M. and Burduroglu, R. (2000), Measuring and selling the value of logistics, The International Journal of Logistics Management, 13(2), 1-18.
20. Lukinsky, V. V. , Barykin, S. E. , and Karpunin, S. A. (2012), Models of inventory management based on the integration of financial and material flows in supply chains, Audit and Financial Analysis, 1, 103-113.
21. Mahmood, M. , AL-kubaisy, W. J. , and Al-Khateeb, B. (2019), Using artificial neural network for multimedia information retrieval, Journal of Southwest Jiaotong University, 54(3), Available from: http://jsju.org/index.php/journal/article/view/296.
22. Mazur, V. V. , Barmuta, K. A. , Demin, S. S. , Tikhomirov, E. A. , and Bykovskiy, M. A. (2016), Innovation clusters: Advantages and disadvantages, International Journal of Economics and Financial Issues, 6(1S), 270-274.
23. Muhammad, K. (2018), The effects of electronic human resource management on financial institutes, Journal of Humanities Insights, 02(01), 1-5.
24. Novikova, N. V. , Barmuta, K. A. , Kaderova, V. A. , Il’Yaschenko, D. P. , Abdulov, R. E. , and Aleksakhin, A. V. (2016), Planning of new products technological mastering and its influence on economic indicators of companies, International Journal of Economics and Financial Issues, 6(8 Special Issue), 65-70.
25. Otter, C. , Watzl, C. , Schwarz, D. , and Priess, P. (2017), Towards sustainable logistics: Study of alternative delivery facets, Entrepreneurship and Sustainability Issues, 4(4), 460-476.
26. Quinn, J. B. (2000), Outsourcing innovation: The new engine of growth, Available from: https://sloanreview.mit.edu/article/outsourcing-innovation-the-new-engine-of-growth/.
27. Rustamova, I. (2013), Conceptual foundations of service industries intangible assets’ management, Journal of Contemporary Economics Issues, 1, Available from: http://economic-journal.net/index.php/CEI/article/view/22.
28. Salih, S. Q. (2019), A new training method based on black hole algorithm for convolutional neural network, Journal of Southwest Jiaotong University, 54(3), Available from: http://jsju.org/index.php/journal/article/view/304.
29. Sharif, A. and Butt, H. (2017), Online businesses and influence of e-marketing on customer satisfaction, Journal of Humanities Insights, 1(2), 89-93.
30. Shestoperov, A. M. (2014), Outsourcing in private and state organizations, Proceedings of the 1st Scientific and Practical Conference Dedicated to the Memory of Honored Professor of Moscow University M. I. Panov, Moscow, Russia, 98-104.
31. Shukla, S. (2017), Innovation and economic growth: A case of India, Humanities & Social Sciences Reviews, 5(2), 64-70.
32. Tyapukhin, A. (2013), Coding and graphic interpretation of the parameters of the logistics flows, Journal of Contemporary Economics Issues, 4, Available from: http://economic-journal.net/index.php/CEI/article/view/86.
33. Zhatkin, D. (2018), Russian literary-critical reception of Burns at turning of the XIX–XX centuries, Opción, 34(85-2), 277-300.
34. Zitkiene, R. and Dude, U. (2018), The impact of outsourcing implementation on service companies, Enterpreneurship and Sustainability Issues, 6(1), 342-355.