Dijana Oreski

Hybrid system with genetic algorithm and artificial neural networks and its application to retail credit risk assessment

The databases of the banks around the world have accumulated large quantities of information about clients and their financial and payment history. These databases can be used for the credit risk assessment, but they are commonly high dimensional. Irrelevant features in a training dataset may produce less accurate results of classification analysis. Data preprocessing is required to prepare the data for classification to increase the predictive accuracy. Feature selection is a preprocessing technique commonly used on high dimensional data and its purposes include reducing dimensionality, removing irrelevant and redundant features, facilitating data understanding, reducing the amount of data needed for learning, improving predictive accuracy of algorithms, and increasing interpretability of models. In this paper we investigate the extent to which the total data, owned by a bank, can be a good basis for predicting the borrowers ability to repay the loan on time. We propose a feature selection technique for finding an optimum feature subset that enhances the classification accuracy of neural network classifiers. Experiments were conducted on the credit dataset collected at a Croatian bank to assess the accuracy of our technique. We found that the hybrid system with genetic algorithm is competitive and can be used as feature selection technique to discover the most significant features in determining risk of default.

Monte-Carlo randomized algorithm: Empirical analysis on real-world information systems

New national and regional bryophyte records, 31 L T Ellis, A Alegro, H Bednarek-Ochyra, R Ochyra, A Bergamini, A Cogoni, P Erzberger, P Górski, N Gremmen, H Hespanhol, C Vieira, L E Kurbatova, M Lebouvier, A Martinčič, A K Asthana, R Gupta, V Nath, R Natcheva, A Ganeva, T Özdemir, N Batan, V Plášek, R D Porley, M Randić, J Sawicki, W Schroder, C Sérgio, V R Smith, P Sollman, S Ştefănuţ, C R Stevenson, G M Suárez, B Surina, G Uyar, Z Modrič Surina The Natural History Museum, UK, University of Zagreb, Croatia, Polish Academy of Sciences, Poland, Swiss Federal Research Institute WSL, Switzerland, Università degli Studi di Cagliari, Italy, Berlin, Germany, Poznań University of Life Sciences, Poland, Diever, The Netherlands, Universidade do Porto, Portugal, Russian Academy of Sciences, Russia, Université de Rennes 1, France, 12 Ljubljana, Slovenia, CSIR-National Botanical Research Institute, India, Bulgarian Academy of Sciences, Bulgaria, Karadeniz Technical University, Turkey, University of Ostrava, Czech Republic, Cerca dos Pomares, Portugal, Public Institution, Croatia, University of Warmia and Mazury in Olsztyn, Poland, Ludwigsstadt, Germany, Universidade de Lisboa, Portugal, University of Stellenbosch, South Africa, St Anna Parochie, The Netherlands, Institute of Biology Bucharest of Romanian Academy, Romania, Norfolk, UK, Facultad de Ciencias Naturales, Argentina, University of Primorska, Slovenia, Zonguldak Karaelmas University, Turkey, Natural History Museum Rijeka, CroatiaDetermination of development priority of information system subsystems is a problem that warrants resolution during information system development. It has been proven, previously, that this problem of information system development order is in fact NP-complete, NP-hard, and APX-hard. To solve this problem on a general case we have previously developed Monte-Carlo randomized algorithm, calculated complexity of this algorithm, and so on. After previous research we were able to come into possession of digraphs that represent real-world information systems. Therefore, in this paper we will empirically analyze Monte-Carlo algorithm to determine how the algorithm works on real-world examples. Also, we will critically review the results and give some possible areas of future research as well.

Effects of dataset characteristics on the performance of feature selection techniques

Display Omitted We connect data characteristics with feature selection techniques performance.Comparative analysis was extensive and included 1280 analysis on 128 data sets.We propose rules for techniques selection based on data characteristics. While extensive research in data mining has been devoted to developing better feature selection techniques, none of this research has examined the intrinsic relationship between dataset characteristics and a feature selection techniques performance. Thus, our research examines experimentally how dataset characteristics affect both the accuracy and the time complexity of feature selection. To evaluate the performance of various feature selection techniques on datasets of different characteristics, extensive experiments with five feature selection techniques, three types of classification algorithms, seven types of dataset characterization methods and all possible combinations of dataset characteristics are conducted on 128 publicly available datasets. We apply the decision tree method to evaluate the interdependencies between dataset characteristics and performance. The results of the study reveal the intrinsic relationship between dataset characteristics and feature selection techniques performance. Additionally, our study contributes to research in data mining by providing a roadmap for future research on feature selection and a significantly wider framework for comparative analysis.

Prediction of academic performance using discriminant analysis

In this paper, discriminant analysis is used as a means of analysing the effect of 30 variables upon the dependent variable Student success at the Faculty of Organization and Informatics, University of Zagreb. The data were collected by a questionnaire administered on two occasions: in the academic year 2006/07, among second-year students of the undergraduate study programme Information and Business Systems, and in the academic year 2008/09, among third- and fourth-year students of the undergraduate study programme Information Systems. This research is aimed at determining predictor variables for student success at the Faculty of Organization and Informatics. The number of students included in the first and second questionnaire administration was 110 and 113, respectively.

Fuzzy knowledge‐based system for calculating course difficulty based on student perception

This research aims to develop a knowledge‐based system used for calculating course difficulty and producing appropriate learning strategies for students. The system is based on fuzzy reasoning and attempts to contribute to the personalization of the learning process. After the description of the data collection process and the search for regularities in the data, we present the service interface. The knowledge‐based system proposed in this paper can yield a number of benefits for both students and institutions: it motivates students to adopt a more practical and personalized approach towards learning and provides an independent learning procedure for both the student and the instructor.

Data-driven decision-making in classification algorithm selection

Abstract The selection of the appropriate classification algorithm for a given data-set is an important and complex issue, full of research challenges. In this paper, we present a developed meta-analysis-based framework to improve decision-making in the selection of classification algorithms based on data-set characteristics. We study the effectiveness of our proposed framework with 32 data-sets. Three classification algorithms – neural networks, decision trees, and k-nearest neighbours – were trained and applied to data-sets with different characteristics, aiming to review the performance of algorithms in the presence of noise in the data, the interaction between features, as well as a small or a large ratio between the number of instances and the number of features. Our results show that feature noise is the most important predictor of the decision regarding the choice of the classification algorithm, and data-driven classification is found to be useful in this scenario.

Handling Sparse Data Sets by Applying Contrast Set Mining in Feature Selection

A data set is sparse if the number of samples in a data set is not sufficient to model the data accurately. Recent research emphasized interest in applying data mining and feature selection techniques to real world problems, many of which are characterized as sparse data sets. The purpose of this research is to define new techniques for feature selection in order to improve classification accuracy and reduce the time required for feature selection on sparse data sets. The extensive comparison with benchmarking feature selection techniques on 64 sparse data sets was conducted. Results have shown superiority of contrast set mining techniques in more than 80% of the analysis on sparse data sets. This paper provides a study on the new methodologies and detected superiority in handling data sparsity.