Is there relationship between cultural-economic indicators and the scientific status of countries? Analysis of Western and Central Asian countries using a neural network algorithm

In this research, the relationship between cultural, economic, R&D indicators with the position and influence of scientific productions of the group countries of India, Turkey, Iran, Saudi Arabia, and Pakistan in Central and Western Asia has been identified. The time period of conducting is from 2001 to 2020, when the required the data were obtained from UNESCO, Scimago, and JCR, and MATLAB, Excel, and also Feed forward Neural Network Algorithm were used for data analysis. The findings show the R&D indicators, economic indicators, and cultural indicators have played a role in improving the scientific position of this group of countries, respectively.


Background and Purpose
The relationship between cultural-economic indicators and STI indicators can somehow show the level of acceptance of each nation about the specific inputs of STI and provide the possibility of evaluating its potential capability in creating the capacity of STI. In fact, the inputs and outputs of countries are influenced by their specific cultural and economic conditions and issues. Accordingly, conducting evaluative analyzes on the development and productivity of countries can be used in line with the goals of decision-making and science policy (Gantman, 2012). Based on this, governments and international organizations always put the measurement and evaluation of STI on their agenda by using different methods and different dimensions (Noroozi Chakoli & Hassanzadeh, 2010;Noroozi Chakoli, 2012).
In this regard, we can refer to the studies of Barro (1991), andVan Raan (2001), each of whom has analyzed various aspects of this issue in their research. INÖNÜ (2003) showed the factors influencing the scientific research output of a country can be classified into two categories of economic and non-economic factors. The first category may be summarily represented by the figures of income per capita or GDP per capita, preferably on the basis of purchasing power parity, for each country while there is no single representation index for the second category. The noneconomic factors bring the influence of the education system, the historical tradition, the science policy of the government, private firms, and other similar sources. All these causes may be related to the cultural background of the country and may be roughly denominated as cultural factors. However, INÖNÜ (2003) emphasized the relationship between the scientific productivity of a country with its GDP or its national income is certainly a very complex one, and hence it is not easy to interpret the meaning of the above classification. Furthermore, Uzun (2006) evaluated Turkey's S&T policy in the last two decades by using various indicators of STI and studied national trends in inputs for R&D activities, publication output, and patent data for the implications of the S&T policy from 1983 to 2003. In the other study, Vinkler (2008) found for EUJ countries correlation between the GDP and the number of publications in a given year proved to be nonsignificant. His longitudinal studies showed, however, significant correlations between the yearly values of GDP and the number of papers published. Studying data referring to consecutive time periods revealed that there is no direct relationship between the GDP and information production of countries. It may be assumed that grants for R&D do not actually depend on real needs, but the fact is that rich countries can afford to spend more whilst poor countries only less money on scientific research. Also, Gontman (2012) examined the influence of economic, linguistic, and political factors on the scientific productivity of countries across selected scientific disciplines. Using a negative binomial regression model, he showed that the effect of these determinants is contingent upon the scientific field under analysis. The only variable that exerts a positive and significant effect across all disciplines is the size of the economy. The linguistic variable only has a positive influence in the social sciences as well as in medicine and agricultural sciences. In addition, it also demonstrated that the degree of political authoritarianism has a negative and statistically significant effect in some of the selected fields. Among other research that analyzed the relationship between two or more variables, it can be mentioned the research of Hossani et al. (2015) examined the association between the use of KM tools and research productivity among four state universities in New York using analysis of variance techniques and found a significant association.
The use of neural network algorithms in scientometric studies has increased significantly in recent years. However, Xu et al. (2022) conducted a scientometric analysis to comprehensively analyze related articles retrieved from seven selected authoritative journals published between 2000 and 2020. They identified co-authorship networks, collaboration networks of countries/regions, co-occurrence networks of keywords, and timeline visualization of keywords, together with the strongest citation burst, the active research authors, countries/regions, and main research interests, as well as their evolution trends and collaborative relationships in the past 20 years. Using the neural network analysis, Daradkeh et al. (2022) divided three dimensions, namely publication features, author features, and content features, into explicit and implicit features to form a set of scientometric terms through explicit feature extraction and implicit feature mapping. The empirical results show that the scientometric classification model proposed in this study performs better than comparable machine learning classification methods in terms of precision, recognition, and F1-score. It also exhibits higher accuracy than deep learning classification based solely on explicit and dominant features.
All these previous studies raise the question in mind, what relationship can be established between cultural-economic characteristics and scientific position? And to what extent have these relations been influential in the scientific position of Western and Central Asian countries? Answering these questions through the communication between multiple indicators using the neural network algorithm can increase the accuracy of the findings and put countries on the right path of S&T policy and lead to the promotion of their S&T status.

2-1. Method, Scope and tools
This research is a type of applied scientometric research that has been conducted crosssectionally and using evaluative techniques, and it has investigated the relationship between cultural-economic indicators and the scientific status of countries using the neural network algorithm. The polulation of this research includes the group countries of India, Turkey, Iran, Saudi Arabia, and Pakistan, which, based on the data of the twenty-year period from 2001 to 2020, had a h-index above the average. The required data were obtained from UNESCO, Scimago, and JCR, and MATLAB and Excel were used for data analysis 1 .
To determine the validity, after designing the neural network, the value of the regression number of at least 0.9 was obtained for the validation graphs of the neural network, which means the acceptable validity of the neural network. In order to review the reliability, the research data was divided into two separate parts of training and testing, so that in one step, the algorithms learn and discover knowledge using the training data, and then, using the set of data Test the discovered knowledge separately. After designing the neural network, the regression diagram of the test data was drawn under the name of network test diagram, and this diagram expressed the level of trust or reliability of the designed neural network.
What is meant by the scientific status of the countries is the status of the countries in the indicators related to the production of science and its influence of the countries. Considering the possibility of research inputs affecting the scientific outputs and productivity of countries after several years, the research data was categorized into four five-year periods based on the period from 2001 to 2020, and data collection was done in the four five-year period. In the next step, the average data of each of the input and output indicators was calculated in a 5-year period for each country.
The reasons for using the Feed forward Neural Network Algorithm for data analysis are the high predictive power of this method, the absence of a specific mathematical relationship between the cultural-economic indicators of the countries and their scientific productivity, the ability to simultaneously predict several variables and the ability of the neural network. It was used to identify important factors through sensitivity analysis (Fausett, 1994;Xu et al. 2022). On the other hand, by increasing the number of inputs and outputs of a problem, discovering the relationship between the output matrix and the input matrix will be more complicated. In such a situation, using common methods such as drawing a graph of output changes according to input, extracting input to output conversion function or other methods is considered impossible or with a very high error. Using neural network as an input-output communicator is an effective approach to solve this problem. A look at the logic of the neural network and the background of using this algorithm confirms the validity of the neural network in solving similar problems and discovering the relationship between inputs and outputs.

2-2. Preparation of neural network input and output data
In the construction of the neural network, normalization is necessary. The data normalization algorithm of this research is that the average of each index in each category was calculated and all the data of the same index for the countries was divided by the calculated average value. With this operation, the entire scale of the input and output data was homogenized. An example of the calculation code of this algorithm in the MATLAB, for the data of the group of countries under study, is given in Figure 1. The matrices a(i) and b(j) are the average matrix of each index for the input and output data, respectively: Based on the categorized data, a neural network was designed for each statistical population and the data related to the indicators of science production and influence were introduced to the neural networks as target variables. Then, all cultural-economic indicators were introduced as input variables to neural networks. Neural networks were implemented with different layers and neurons, and after training, validating and testing the neural network model, the performance of the networks was compared based on the matching of the output data with the target data in the test and validation, and the network that had the best performance was used to identify Important factors were used using sensitivity analysis .

2-3. Design and construction of neural network
After preparing the input and output data, the neural network was designed and built. To discover the relationship between the input matrix and the output matrix and design the network using Feed forward Neural Network Algorithm, the neural network toolbox figure 2 was used in the MATLAB. After calling the toolbox, in the first step, input data and output data (target) were defined for the network (Figure 3). Figure 3. Definition of input and output data matrix for neural network In the second step, the desired neural network type was defined and the number of layers and neurons were determined. The transfer function for the output of the neurons is also one of the other variables that should be known at this stage. Figure 4 shows how to build a neural network and the available variables for network design by the toolbox. Changing each of the variables will lead to the construction and design of the neural network with its own characteristics. After determining the parameters and variables, the neural network was built, whose overall structure (number of layers and neurons in each layer) is displayed by MATLAB ( Figure 5). As seen in figure 5, the constructed neural network consists of 2 layers (a hidden layer and an output layer), with the middle layer having 8 neurons and the output layer having 9 neurons. After constructing the neural network, network training was done with input and target data. To start training the network and stop it, separate parameters are defined as criteria for stopping the training of the network. After defining the training parameters of the network, the neural network starts training with the input data and the defined goal ( Figure 6). The process of training the network continues until at least one of the stopping criteria of the network is satisfied. Figure 6. Neural network training process Further, in order to analyze the research data, the neural network design and construction algorithm was implemented for all the countries of India, Turkey, Iran, Saudi Arabia and Pakistan with their input data and output, and finally the neural network of each the category that has the ability to convert input matrix (inputs) to output (outputs) was obtained.
• Input indicators: -Cultural indicators (including: literacy rate, illiterate population, number of incoming international students, number of outgoing international students, spending on education as a percentage of GDP, spending on education as a percentage of total government spending); -Economic indicators (including: GDP, GDP adjuster, GDP growth, GDP per capita, total government expenditure, and per capita gross income of the country);

2-4. Characteristics of the neural network of the studied countries
The neural network designed for the studied countries has 3 layers, two of which are hidden and each has 6 neurons, and the last layer has 9 neurons as the output layer (Figure 7).

3-1. Analyzing the relationship between cultural indicators and indicators of science production and the influence
At this stage, first, the relationship between cultural indicators and output indicators was studied using the neural network algorithm, and the results are presented in tables 1 to 5. As can be seen in these tables, the literacy rate has a positive significant relationship with the number of scientific productions, and a negative significant relationship with the citation ratio of each scientific document; Illiterate population index has a significant negative relationship with number of scientific productions, total number of citations of scientific documents and share of international scientific production; the number of international student output has a significant positive relationship with the h-index and number of journals indexed in JCR and does not have a significant relationship with other indicators of scientific production; The expenditure in education as a percentage of GDP has no significant relationship with most indicators of scientific production and has a significant negative relationship with the citation rate of each scientific document and IF of journals indexed by JCR" and with the number Journals indexed in JCR have a positive significant relationship; The index expenditure in education as a percentage of total government expenditure has no significant relationship with any of the indicators of scientific production. Table 1. The effect of changing the input of the literacy rate on the output Table 2. The effect of changing the input of the illiterate population on the output Table 3. The effect of changing the input and output of international students on the output Table 4. The effect of changing the input of spending on education as a percentage of GDP on the output Table. 5. The effect of changing the input of expenditure on education as a percentage of total government expenditure on the output

3-2. Analysis of the relationship between economic indicators and the indicators of scientific production and the influence
The relationship between six economic indicators, including GDP, GDP adjuster, GDP growth, GDP per capita, total government expenditure, and per capita gross income of the country with output indicators (science production and the influence) in the countries in Tables 6 Up to 11 are shown. As can be seen, the GDP index has a positive significant relationship with indicators such as "total number of published scientific documents", "total number of citations on scientific documents" and "share of international science production", but with indicators such as the ratio of citations per scientific document and h-index have a significant negative relationship; The GDP adjusting indicator does not have a significant relationship with the indicators of scientific production and the influence; GDP Growth has a positive significant relationship with indicators such as H and the number of journals in the JCR, but has a negative significant relationship with indicators such as the total number of citations on scientific documents and the indictor of the IF of journals in the JCR; The GDP per capita indicator has a positive significant relationship with indicators such as H and the number of indexed journals in the JCR, but it has a negative significant relationship with the indicator of the IF of the journals in the JCR; The indicator of the total public expenditure of the government has a positive significant relationship with the indicators of the total number of published scientific documents, the share of international scientific production and the IF, and a negative significant relationship with the indicators of the ratio of citations per scientific document, H and the number of journals indexed by JCR; and the per capita index of gross income has no significant relationship with the indicators of scientific production and its influence. Table 6. The effect of changes in GDP input on the output Table 7. The effect of changing the adjusting input of GDP on the output Table 8. The effect of the change in the adjusting input of GDP on the output Table 9. The effect of changes in GDP per capita on the outputs Table 10. The effect of changes in the total input of public expenditure of the government on the output Table 11. The effect of changes in per capita gross income on the output

3-3. Analysis of the relationship between R&D indicators and science production indicators and the influence
The relationship between six R&D indicators including the number of R&D personnel, the number of researchers, the number of technicians and equivalent employees, the number of other R&D employees, the total gross domestic expenditure on R&D and the percentage of the total gross domestic expenditure on R&D from the GDP Domestic with indicators of science production and its influence in the Central and Western Asian countries are specified in tables 12 to 17. As can be seen, the indicators of the number of R&D personnel, the number of researchers and the number of other R&D employees have a significant positive relationship with the indicators of the total number of scientific papers published, the total number of citations, the share of international science production and the IF of journals indexed by JCR; In contrast, the indicators of the number of technicians and equivalent employees and the total gross domestic expenditure of R&D are completely opposite to the previous indicators and have the opposite effect. In other words, there is a significant negative relationship with the indicators of the total number of scientific documents published, the total number of citations, the share of international science production and the IF of journals indexed by JCR; And the index of the percentage of the total gross domestic expenditure of R&D from the GDP has a positive significant relationship with the total number of published scientific documents and a non-significant relationship with other indicators.  Table 13. The effect of changing the input of the number of researchers on the output Table 14. The effect of changing the input of the number of technicians and equivalent employees on the output Table 15. The effect of changing the input of the number of other R&D employees on the output Table 16. The effect of changing the input of the total gross domestic expenditure of R&D on the output Table 17. The effect of changing the input percentage of the total gross domestic expenditure of R&D from the gross domestic product on the output

Discussion
The neural network algorithm allows the simultaneous analysis of the relationship of multiple indicators and can lead the measurement of research productivity to become more qualitative. According to the findings of this research, in the Central and Western Asian countries, the literacy rate has a significant positive relationship with the number of scientific productions and a negative significant relationship with the ratio of citations on each scientific document; Illiterate population index has a significant negative relationship with the number of scientific productions, the total number of citations and the share of international science production. This means that as the illiterate population decreases, scientific productivity increases. The findings indicate that the GDP has a significant positive relationship with the total number of scientific documents, the total number of citations and the share of international science production, but with indicators such as the ratio of citations on each scientific document and the h-index has a negative significance. This means that the GDP of the West and Central Asian countries has a significant positive relationship with the quantitative indicators of scientific production and a negative significant relationship with the qualitative indicators.
R&D indicators, including the number of R&D personnel, the number of researchers and the number of other R&D employees, have a significant positive relationship with the total number of published scientific documents, the total number of citations, the share of international scientific production, and the IF of the journals indexed by JCR. This means that by investing in this regard, the scientific position of the countries can be developed.  Inönü (2003), Uzun (2006), and show that there is a significant relationship between educational, economic, and R&D situation with scientific position and influenc of the countries.

Conclusions
According to the findings, In order to increase the share of international science production and increase scientific production, special focus and attention should be paid to the indicators of the number of R&D personnel, the number of other R&D employees, GDP, the number of researchers and the total public expenditure of the government. The h-index, which shows the influence of scientific productions, has been affected by the number of technicians and equivalent employees and the growth of GDP more than any other index. Based on the findings, R&D indicators, economic indicators, and cultural indicators have played a role in improving the scientific position of this group of countries, respectively.