Nataša Šarlija

Modelling small-business credit scoring by using logistic regression, neural networks and decision trees

Previous research on credit scoring that used statistical and intelligent methods was mostly focused on commercial and consumer lending. The main purpose of this paper is to extract important features for credit scoring in small-business lending on a dataset with specific transitional economic conditions using a relatively small dataset. To do this, we compare the accuracy of the best models extracted by different methodologies, such as logistic regression, neural networks (NNs), and CART decision trees. Four different NN algorithms are tested, including backpropagation, radial basis function network, probabilistic and learning vector quantization, by using the forward nonlinear variable selection strategy. Although the test of differences in proportion and McNemars test do not show a statistically significant difference in the models tested, the probabilistic NN model produces the highest hit rate and the lowest type I error. According to the measures of association, the best NN model also shows the highest degree of association with the data, and it yields the lowest total relative cost of misclassification for all scenarios examined. The best model extracts a set of important features for small-business credit scoring for the observed sample, emphasizing credit programme characteristics, as well as entrepreneurs personal and business characteristics as the most important ones. Copyright

The impact of liquidity on the capital structure: a case study of Croatian firms

The impact of liquidity on the capital structure: a case study of Croatian firms Background: Previous studies have shown that in some countries, liquid assets increased leverage while in other countries liquid firms were more frequently financed by their own capital and therefore were less leveraged. Objectives: The aim of this paper is to investigate the impact of liquidity on the capital structure of Croatian firms. Methods/Approach: Pearson correlation coefficient is applied to the test on the relationship between liquidity ratios and debt ratios, the share of retained earnings to capital and liquidity ratios and the relationship between the structure of current assets and leverage. Results: A survey has been conducted on a sample of 1058 Croatian firms. There are statistically significant correlations between liquidity ratios and leverage ratios. Also, there are statistically significant correlations between leverage ratios and the structure of current assets. The relationship between liquidity ratios and the short-term leverage is stronger than between liquidity ratios and the long-term leverage. Conclusions: The more liquid assets firms have, the less they are leveraged. Long-term leveraged firms are more liquid. Increasing inventory levels leads to an increase in leverage. Furthermore, increasing the cash in current assets leads to a reduction in the short-term and the long-term leverage.

Shaping the Entrepreneurial Mindset: Entrepreneurial Intentions of Business Students in Croatia

Business students from a public university in roatia participated in an international study on entrepreneurial self‐efficacy, identity, and education. The results of this preliminary empirical research indicate that the main predictors of the entrepreneurial intentions in roatia are strength of entrepreneurial identity aspiration and entrepreneurial self‐efficacy. These two main constructs mediate the number of personal, situational, or contextual factors, including education. Empirical analysis supports the majority of ocial ognitive areer heory hypothesized interaction between control variables and main constructs such as self‐efficacy, positive outcome expectations, and entrepreneurial identity. These findings thus fill the gap in the empirical evidence of the theoretical framework validity derived from different contexts.

Gender Role and Religion as Predictors of Attitude Toward Abortion in Croatia, Slovenia, the Czech Republic, and the United States

Since socialism’s decline, abortion has become a divisive political issue in East Central Europe, just as it is in the United States. Questionnaires administered to college students in Croatia, the Czech Republic, Slovenia, and the United States between 1991 and 1994 assessed the importance of religious identification, degree of religious feelings and participation, desired number of children, and gender role attitudes as predictors of approval of abortion for reasons of personal choice. Multiple regression indicated that these variables best predicted level of approval of abortion in Slovenia. The strong link between gender role attitude and abortion approval that emerged in the Slovene 1993 data is discussed in relation to the changing social and political contexts.

The Relationship between Entrepreneurial Activities, National and Regional Development and Firm Efficiency—Global Entrepreneurship Monitor (GEM)-based Evidence from Croatia

This study analyses the dynamics, structure and connections between entrepreneurial activity, economic development and firm efficiency. While the usual presumption on the relationship between these variables implies straightforward, linear and positive impacts, empirical evidence shows that those impacts are significant, more complex and less straightforward. The evidence of entrepreneurial activity in Croatia shows that the early stages of entrepreneurship development are very dynamic and volatile. Furthermore, significant inter-regional differences exist in entrepreneurial activity, firm performance and economic development across six Croatian regions. Correlations between entrepreneurial activity, firm performance and economic development are significant but depend on whether the entrepreneurial activity is opportunity or necessity based. This study confirms the theoretical presumption about complex and multilayered connections between different types of entrepreneurship activity and economic development.

A Comparison of Machine Learning Methods in a High-Dimensional Classification Problem

Abstract Background: Large-dimensional data modelling often relies on variable reduction methods in the pre-processing and in the post-processing stage. However, such a reduction usually provides less information and yields a lower accuracy of the model. Objectives: The aim of this paper is to assess the high-dimensional classification problem of recognizing entrepreneurial intentions of students by machine learning methods. Methods/Approach: Four methods were tested: artificial neural networks, CART classification trees, support vector machines, and k-nearest neighbour on the same dataset in order to compare their efficiency in the sense of classification accuracy. The performance of each method was compared on ten subsamples in a 10-fold cross-validation procedure in order to assess computing sensitivity and specificity of each model. Results: The artificial neural network model based on multilayer perceptron yielded a higher classification rate than the models produced by other methods. The pairwise t-test showed a statistical significance between the artificial neural network and the k-nearest neighbour model, while the difference among other methods was not statistically significant. Conclusions: Tested machine learning methods are able to learn fast and achieve high classification accuracy. However, further advancement can be assured by testing a few additional methodological refinements in machine learning methods.

Business Client Segmentation in Banking Using Self-Organizing Maps

Abstract Segmentation in banking for the business client market is traditionally based on size measured in terms of income and the number of employees, and on statistical clustering methods (e.g. hierarchical clustering, k-means). The goal of the paper is to demonstrate that self-organizing maps (SOM) effectively extend the pool of possible criteria for segmentation of the business client market with more relevant criteria, including behavioral, demographic, personal, operational, situational, and cross-selling products. In order to attain the goal of the paper, the dataset on business clients of several banks in Croatia, which, besides size, incorporates a number of different criteria, is analyzed using the SOM-Ward clustering algorithm of Viscovery SOMine software. The SOM-Ward algorithm extracted three segments that differ with respect to the attributes of foreign trade operations (import/export), annual income, origin of capital, important bank selection criteria, views on the loan selection and the industry. The analyzed segments can be used by banks for deciding on the direction of further marketing activities.

Gender Patterns of Businesses with Growth Potential in Croatia

The literature review presented in the chapter confirms that researchers’ interest is more focused on why and how businesses are created, and much less why and how they grow. There is a shortage of empirical evidence of growth influencers, as well as evidence about the intensity and patterns of the interaction between internal factors and entrepreneurship ecosystem in the process of growing a venture. The authors identify this gap and combine different angles (theory of firm growth, entrepreneurship, inclusion, macroeconomic aspects of using resources) with a gender aspect (inclusion), in exploring why and how businesses grow. Using this angle and aggregated Global Entrepreneurship Monitor (GEM) data from the 2003–2013 period, the chapter presents several gender patterns of businesses with growth potential (innovative products, innovative technology, and competitiveness) in Croatia.

Dealing with Interpretability Issues in Predicting Firm Growth: Factor Analysis Approach

Abstract Background: Alongside the theoretical progress made in understanding the factors that influence firm growth, many methodological challenges are yet to be overcome. Authors point to the notion of interpretability of growth prediction models as an important prerequisite for further advancement of the field as well as enhancement of models’ practical values. Objectives: The objective of this study is to demonstrate the application of factor analysis for the purpose of increasing overall interpretability of the logistic regression model. The comprehensive nature of the growth phenomenon implies propensity of input data to be mutually correlated. In such situations, growth prediction models can demonstrate adequate predictability and accuracy, but still lack the clarity and theoretical soundness in their structure. Methods/Approach: The paper juxtaposes two prediction models: the first one is built using solely the logistic regression procedure, while the second one includes factor analysis prior to development of a logistic regression model. Results: Factor analysis enables researchers to mitigate inconsistencies and misalignments with a theoretical background in growth prediction models. Conclusions: Incorporating factor analysis as a step preceding the building of a regression model allows researchers to lessen model interpretability issues and create a model that is easier to understand, explain and apply in real-life business situations.