Search

Modeling the cost of energy consumption of public buildings is vital for planning reconstruction measures in the public sector. The methods of predictive analytics have not been sufficiently exploited in this domain. This paper aimed to create a model for predicting the cost of the total energy consumption of a building based on deep learning (DL) and compare it to the standard linear regression (MLR), as well to identify key predictors that can significantly influence the cost of energy. An algorithm for modeling procedure is proposed which includes data pre- processing, variable reduction procedures, training and testing MLR and deep neural networks (DNN) and, finally, performance evaluation. Variable reduction in the MLR model was conducted by a backward procedure ; while in DNN, the Olden method was used. The algorithm was tested on a high-dimensional real dataset of Croatian public buildings. The results showed that there is a statistically significant difference in the distribution of DNN predictions and distribution of actual values in the validation set, as opposed to distribution of MLR predictions and real values. However, DNN model had a lower normalized root mean square error, while the MLR model had a lower symmetric mean average error. Those findings reveal the potential of DL for solving this type of problems but also the need for more advanced algorithms adjusted to deal with large-range numeric outputs. The created models could be implemented in public sector business intelligence systems to support policy and decision makers in allocating resources for building reconstructions.

The paper deals with modeling the cost of energy consumed in public buildings by leveraging three machine learning methods: artificial neural networks, CART, and random forest regression trees. Energy consumption is one of the major issues in global and national policies, therefore scientific efforts in creating prediction models of energy consumption and cost are highly important. One of the largest energy consumers in every state is its public sector, consisting of educational, health, public administration, military, and other types of public buildings. Recent technologies based on sensor networks and Big data platforms enable collection of large amounts of data that could be used to analyze energy consumption and cost. A real data from Croatian public sector is used in this paper including a large number of constructional, energetic, occupational, climate and other attributes. The algorithms for data pre-processing and modeling by optimizing parameters are suggested. Three strategies were tested: (1) with all available variables, (2) with a filter-based variable selection, and (3) with a wrapper-based variable selection which integrates Boruta algorithm and random forest. Prediction models of energy cost are created using two approaches: (a) comparative usage of artificial neural networks and two types of regression trees, CART and random forest, and (b) integration of RF-Boruta variable selection and machine learning methods for prediction. A cross-validation procedure was used to optimize the artificial neural network and regression tree topology, as well to select the most appropriate activation function. Along with creating a prediction model, the aim of the paper was also to extract the relevant predictors of energy cost in public buildings which are important in planning the construction or renovation of buildings. The results have shown that the second approach which integrates machine learning with Boruta method, where the random forest algorithm is used for both variable reduction and prediction modeling, has produced a higher accuracy of prediction than the individual usage of three machine learning methods. Such findings confirm the potential of hybrid machine learning methods which are suggested in previous research, but in favor of random forest method over CART and artificial neural networks. Regarding the variable selection, the model has extracted heating and occupational data as the most important, followed by constructional, cooling, electricity, and lighting attributes. The model could be implemented in public buildings information systems and their IoT networks within the concept of smart buildings and smart cities.

A new teaching approach is presented which integrates observational learning through field teaching of business processes and simulation modeling in order to increase students' learning outcomes and acceptance of computer simulation technology. The teaching method, called LOSI (learning by observing, simulating, and improving), was conducted at a Croatian high education institution. The efficiency of the LOSI approach was investigated by conducting a survey based on the technology acceptance model (TAM). The indicators of ease of use, usefulness, and enjoyment in participating in LOSI were collected along with students' grades and their intention to use this technology in future work and education. The inter-relations among variables were analyzed by statistical tests. The results revealed that students find LOSI easy to use, useful in achieving learning outcomes, and highly enjoyable, while the ease of use and enjoyment is positively associated to usefulness (i.e., learning outcomes).

Energy efficiency of public sector is an important issue in the context of smart cities due to the fact that buildings are the largest energy consumers, especially public buildings such as educational, health, government and other public institutions that have a large usage frequency. However, recent developments of machine learning within Big Data environment have not been exploited enough in this domain. This paper aims to answer the question of how to incorporate Big Data platform and machine learning into an intelligent system for managing energy efficiency of public sector as a substantial part of the smart city concept. Deep neural networks, Rpart regression tree and Random forest with variable reduction procedures were used to create prediction models of specific energy consumption of Croatian public sector buildings. The most accurate model was produced by Random forest method, and a comparison of important predictors extracted by all three methods has been conducted. The models could be implemented in the suggested intelligent system named MERIDA which integrates Big Data collection and predictive models of energy consumption for each energy source in public buildings, and enables their synergy into a managing platform for improving energy efficiency of the public sector within Big Data environment. The paper also discusses technological requirements for developing such a platform that could be used by public administration to plan reconstruction measures of public buildings, to reduce energy consumption and cost, as well as to connect such smart public buildings as part of smart cities. Such digital transformation of energy management can increase energy efficiency of public administration, its higher quality of service and healthier environment.

The cost of energy consumed in educational, health, public administration, military, and other types of public buildings constitutes a substantial proportion of the total expenditure of the public sector. Due to a large number of attributes that influence the energy cost of a building, most of the models developed in the literature use only a subset of predictors, often neglect occupational data, and do not exploit enough the potential of deep learning methods. In this paper a real data from Croatian public sector is used including constructional, energetic, geographical, occupational and other attributes. Algorithms for data preprocessing and for deep learning modelling procedure are suggested. The number of hidden units in the deep neural network is optimized by a cross-validation procedure, while the sigmoid activation function was tested with Adam optimization algorithm. The feature selection was conducted using the recursive feature elimination method with a regression random forest kernel. The aims were to identify the subset of relevant predictors of energy cost in public buildings that could assist decision makers in determining the priority of reconstruction measures as well as to test the potential of deep learning in predicting the yearly energy cost. The results have shown that the deep learning network with three hidden layers was the most successful in predicting energy cost using the wrapper-based method of feature extraction. The selection of features confirms the importance of occupational data, as well as heating, cooling, electricity lightning, and constructional attributes for estimating the total energy cost. Those predictors can be used in decision making on allocating resources in public buildings reconstructions. The model implementation could improve public sector energy efficiency, save costs and contribute to the concepts of smart buildings and smart cities.

Scientific efforts in predicting energy efficiency and consumption align with the European Commission directives about reducing greenhouse gas emissions, increasing energy efficiency and using 20% of energy from renewable resources until 2020. The largest individual energy consumer is the building sector which contains 40% of total primary energy consumption (Tommerup et al., 2007). That stresses out the importance of creating efficient models that will be able to extract features and predict the energy efficiency level of a building according to its characteristics and planned reconstruction measures. This work deals with creating such prediction model by using data from Croatian Agency for Legal Trade and Real Estate Brokerage (APN) which maintains the centralized information system of energy efficiency in public buildings (ISGE). In order to create prediction models, the machine learning methods were used such as clustering and artificial neural networks. The real dataset of Croatian public buildings with a large number of attributes which were reduced in pre- modelling phase. The aim of this research was to investigate the effect of a clustering procedure on artificial neural network model accuracy. For that purpose, the symmetric mean average percentage error (MAPE) of an initial prediction model obtained on the whole dataset is compared to the MAPE of separate models obtained on each cluster, and the results were discussed. Due to a lack of approaches which integrate clustering and neural networks in modelling energy efficiency, the scientific contribution of this paper is in providing such an approach and in analysing its effects to modeling accuracy. Guidelines for using the model as an approach for saving costs in public sector are also given.

In this paper, we consider the problem of searching for an optimal partition with the most appropriate number of clusters for an incomplete data set in which several outliers might occur. Special attention is given to the application of the Least Squares distance-like function. The procedure of preparing the incomplete data set and the outlier elimination procedure are proposed such that the clustering process gives acceptable solutions. Appropriate justifications with proof are provided for these procedures. An incremental algorithm for searching for optimal partitions with 2, 3, ... clusters is applied on the prepared data set. After that, by using the Davies-Bouldin and the Calinski Harabasz index the most appropriate number of clusters is determined. The whole procedure is organized as an algorithm given in the paper. In order to illustrate its applicability, the above steps are applied on the real data set of public buildings and their energy efficiency data, providing clear clusters that could be used for further modeling procedures.

In an age where information technology, the Internet, smart phones, mobile applications, interdisciplinarity, big data, business analytics, artificial intelligence, and the fourth industrial revolution have become prevalent, Operational Research (OR) as a scientific and academic discipline is also subject to a specific transformation and evolution. This paper covers the current situation with OR based, on the mentioned trends by analysing OR-related programmes and courses offered at university level, specifically, at higher education institutions (HEIs) in Croatia. Also, aspects of collaboration between HEIs and the business community relating to OR will be studied, as well as the role of the Croatian Operational Research Society (CRORS) as a community that gathers OR researchers and promotes OR as a scientific discipline (Croatian Operational Research Society 2017). Some suggestions to improve the current OR situation at Croatian HEIs, and its impact on economy are also given. The analysis was based on a survey sent to HEIs in Croatia, and subsequently organising a round table discussion and consultations with Croatian employers. The survey questions related to the courses and programmes offered at Croatian HEIs, the number of students attending the courses, recent innovations, collaboration with the business community and plans to make OR even more appealing. The results reveal that the number of OR-related courses is significant, though most are electives, and the lack of OR majors across all higher-education levels in Croatia. In general, OR courses are required to follow terminology trends and utilize more appealing terms such as data analytics and business analytics. Recent and more intensive collaboration with the business community which seeks experts in optimization, analytics, and other OR areas is becoming promising.

Association rules are a data mining method for discovering patterns of frequent item sets, such as products in a store that are frequently purchased at the same time by a customer (market basket analysis). A number of interestingness measures for association rules have been developed to date, but research has shown that there a dominant measure does not exist. Authors have mostly used objective measures, whereas subjective measures have rarely been investigated. This paper aims to combine objective measures such as support, confidence and lift with a subjective approach based on human expert selection in order to extract interesting rules from a real dataset collected from a large Croatian retail chain. Hierarchical association rules were used to enhance the efficiency of the extraction rule. The results show that rules that are more interesting were extracted using the hierarchical method, and that a hybrid approach of combining objective and subjective measures succeeds in extracting certain unexpected and actionable rules. The research can be useful for retail and marketing managers in planning marketing strategies, as well as for researchers investigating this field.

Background: Previous research has shown success of data mining methods in marketing. However, their integration in a knowledge management system is still not investigated enough. Objectives: The purpose of this paper is to suggest an integration of two data mining techniques: neural networks and association rules in marketing modeling that could serve as an input to knowledge management and produce better marketing decisions. Methods/Approach: Association rules and artificial neural networks are combined in a data mining component to discover patterns and customers’ profiles in frequent item purchases. The results of data mining are used in a web-based knowledge management component to trigger ideas for new marketing strategies. The model is tested by an experimental research. Results: The results show that the suggested model could be efficiently used to recognize patterns in shopping behaviour and generate new marketing strategies. Conclusions: The scientific contribution lies in proposing an integrative data mining approach that could present support to knowledge management. The research could be useful to marketing and retail managers in improving the process of their decision making, as well as to researchers in the area of marketing modelling. Future studies should include more samples and other data mining techniques in order to test the model generalization ability.

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.

The paper aims to establish an efficient model for predicting company growth by leveraging the strengths of logistic regression and neural networks. A real dataset of Croatian companies was used which described the relevant industry sector, financial ratios, income, and assets in the input space, with a dependent binomial variable indicating whether a company had high-growth if it had annualized growth in assets by more than 20% a year over a three-year period. Due to a large number of input variables, factor analysis was performed in the pre -processing stage in order to extract the most important input components. Building an efficient model with a high classification rate and explanatory ability required application of two data mining methods: logistic regression as a parametric and neural networks as a non -parametric method. The methods were tested on the models with and without variable reduction. The classification accuracy of the models was compared using statistical tests and ROC curves. The results showed that neural networks produce a significantly higher classification accuracy in the model when incorporating all available variables. The paper further discusses the advantages and disadvantages of both approaches, i.e. logistic regression and neural networks in modelling company growth. The suggested model is potentially of benefit to investors and economic policy makers as it provides support for recognizing companies with growth potential, especially during times of economic downturn.

A nonlinear approach for detecting relative lymphopenia is suggested by using a health data record based on simple clinical parameters. Two classification methods, neural networks and decision trees, were applied to detect whether a patient has a positive or a negative lymphopenia outcome. Due to a large dimension of input space, a feature selection method was used in the pre-processing stage. All tested models were validated on the same out-of-sample dataset, and a 10-fold cross-validation procedure for testing generalization ability of the models was conducted. The models were compared according to their classification accuracy in the sense of the average hit rate, specificity and sensitivity. The results show that (1) the best neural network model slightly outperforms the decision tree model, (2) the reduced model provides even higher accuracy than the models with all available data, and (3) both methods similarly rank five important predictors of lymphopenia. The paper discusses the relevance of extracted features, and suggests some guidelines for further research.

In this paper we consider the problem of natural gas consumption hourly forecast on the basis of hourly movement of temperature and natural gas consumption in the preceding period. There are various methods and approaches for solving this problem in the literature. Some mathematical models with linear and nonlinear model functions relating to natural gas consumption forecast with the past natural gas consumption data, temperature data and temperature forecast data are mentioned. The methods are tested on concrete examples referring to temperature and natural gas consumption for the area of the city of Osijek (Croatia) from the beginning of the year 2008. The results show that most acceptable forecast is provided by mathematical models in which natural gas consumption and temperature are related explicitly.

The purpose of the paper was to extract important features of children's mathematical gift by using neural networks and logistic regression, in order to create a model that will assist teachers in elementary schools to recognize mathematically gifted children in an early stage, therefore enabling further development and realization of that gift. The initial model was created on the basis of a theoretical background and heuristical knowledge on giftedness in mathematics, including five components: (1) mathematical competencies, (2) cognitive components of gift, (3) personal components that contribute gift development, (4) environmental factors, and (5) efficiency of active learning and exercising methods, as well as grades and out-of-school activities of pupils in the fourth year of elementary school. The three neural network classification algorithms were tested in order to extract the important variables for detecting mathematically gifted children. The best neural network model was selected on the basis of a 10-fold cross-validation procedure. The model was also investigated by the logistic regression. Important predictors detected by two methods were compared and analyzed. The results show that both methods extract similar set of variables as the most important, including grades in mathematics, mathematical competencies of a child regarding numbers and calculating, but also grades in the literature, and environmental factors.

This paper compares the efficiency of two intelligent methods: expert systems and neural networks, in detecting children's mathematical gift at the fourth grade of elementary school. The input space for the expert system and the neural network model consisted of 60 variables describing five basic components of a child's mathematical gift identified in previous research. The expert system estimated a child's gift based on heuristically defined logic rules, while the scientifically confirmed psychological evaluation of gift based on Raven's standard progressive matrices was used at the output of neural network models. Three neural network algorithms were tested on a Croatian dataset. The results show that both the expert system and the neural network recognize more pupils as mathematically gifted than teachers do. The expert system produces the highest average hit rate, although the highest accuracy in classifying gifted children is obtained by the radial basis neural network algorithm, which also yields lower type II error. Due to the ability of expert systems to explain the result, it can be suggested that both the expert system and the neural network model have potential to serve as effective intelligent decision support tools in detecting mathematical gift in early stage, therefore enabling its further development.

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 McNemar's 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 entrepreneur's personal and business characteristics as the most important ones. Copyright © 2005 John Wiley & Sons, Ltd.

Choice models and neural networks are two approaches used in modeling selection decisions. Defining model performance as the out-of-sample prediction power of a model, we test two hypotheses: (i) choice models and neural network models are equal in performance, and (ii) hybrid models consisting of a combination of choice and neural network models perform better than each stand-alone model. We perform statistical tests for two classes of linear and nonlinear hybrid models and compute the empirical integrated rank (EIR) indices to compare the overall performances of the models. We test the above hypotheses by using data for various brand and store choices for three consumer products. Extensive jackknifing and out-of-sample tests for four different model specifications are applied for increasing the external validity of the results. Our results show that using neural networks has a higher probability of resulting in a better performance. Our findings also indicate that hybrid models outperform stand-alone models, in that using hybrid models guarantee overall results equal or better than the two stand-alone models. The improvement is particularly significant in cases where neither of the two stand-alone models is very accurate in prediction, indicating that the proposed hybrid models may capture aspects of predictive accuracy that neither stand-alone model is capable of on their own. Our results are particularly important in brand management and customer relationship management, indicating that multiple technologies and mixture of technologies may yield more accurate and reliable outcomes than individual ones.

The aim of this paper was to model a neural network capable of detecting mathematically gifted fourth-grade elementary school pupils. The input space consisted of variables describing the five basic components of a child's mathematical gift identified in the body of previous research. The scientifically confirmed psychological evaluation of gift based on Raven's standard progressive matrices was used at the output. Three neural network models were tested on a Croatian dataset: multilayer perceptron, radial basis, and probabilistic network. The models' performances were measuredaccording to the average hit rate obtained on the test sample. According to the results, the highest accuracy is produced by the radial basis neural network, which correctly recognizes all gifted children. Such high classification accuracy shows that neural networks have the potential to serve as an effective intelligent decision support tool able to assist teachers in detecting mathematically gifted children. This can be particularly useful in schools in which there is a shortage of psychologists, Cilj ovoga rada bio je modeliranje neuronske mreže kojom bi se mogla otkriti matematička darovitost u učenika četvrtih razreda osnovnih škola. Ulaz se sastojao od varijabli izvedenih za opis pet osnovnih komponenata matematičke darovitosti u djece, a koje su ustanovljene u prethodnim istraživanjima. Kao izlazni rezultat upotrijebljena je znanstveno potvrđena psihološka evaluacija darovitosti utemeljena u Ravenovim progresivnim matricama. Tri modela neuronskih mreža testirana su na hrvatskim podatcima: višeslojni perceptron, mreža s radijalno zasnovanom funkcijom i probabilistička (vjerojatnosna) mreža. Rad mreža mjeren je u odnosu na prosječnu stopu pogodaka prikupljenih na testnom uzorku. Analiza je pokazala da je najvišu točnost postigla neuronska mreža s radijalno zasnovanom funkcijom, kojom se mogu točno prepoznati sva darovita djeca. Tako visoka točnost u klasifikaciji pokazuje da neuronske mreže imaju potencijal služiti kao efektivan alat inteligentne odluke pomoću kojega bi učitelji mogli otkriti djecu s darovitošću za matematiku. To može biti osobito korisno u školama s manjkom psihologa.

Rad se bavi modeliranjem sustava zasnovanih na znanju s ciljem povećanja kvalitete odlučivanja u segmentu veleprodaje. Ekspertni sustavi, kao tehnika sustava zasnovanih na znanju, računalni su programi koji se služe znanjem eksperta iz nekoga specijaliziranog područja u svrhu potpore odlučivanju. U radu je dan pregled prethodnih istraživanja upotrebe sustava zasnovanih na znanju u prodaji, kao i mogućnosti njihove prilagodbe potrebama i preferencama korisnika. U cilju provjere tih mogućnosti dizajniran je eskpertni sustav za donošenje odluke o odobrenju prodaje robe određenom kupcu u veleprodaji. Predložena baza znanja ekspertnog sustava koristi se produkcijskim pravilima. S pomoću ljuske programa Exsys Corvid izgrađen je ekspertni sustav koji je u testnoj fazi pokazao da je njegovom uporabom moguće ubrzati proces prodaje unutar tvrtke i povećati učinkovitost prodajnog poslovanja. U radu su analizirane neke prednosti i nedostatci uporabe takvih sustava, The paper deals with modeling knowledg based systems with the aim of increasing the quality of decision making in the segment of wholesales. Expert systems, as a techinque of knowledge based systems, are computer programs that use the knowledge from an expert in a specialized area in purpose of decision support. A review of previous research of knowledge based systems in sales is given in the paper, as well as the possibilities of their adjustment to the user needs and preferences. In order to test those abilities, an expert system aimed for decision support in the process of selling goods to a certain customer is designed. The suggested knowledge base uses production rules to represent knowledge. By using the expert system shell ExSys Corvid, an expert system is built which showed in its testing phase that it is possible to make the sale process faster and more efficient. Some benefits and limitations of the suggested system are discussed in the paper.

A scientifically approved psychological finding of gift is usually not available to all schools. In order to obtain an accurate and early detection of mathematically gifted children, an intelligent expert system MathGift is proposed to assist teachers in making decision about a child's gift in mathematics in the fourth grade of elementary school. Besides mathematical competencies, the system includes other components while deciding about the giftedness in mathematics, such as cognitive components of gift, personal components that contribute to gift development, strategies of learning and exercising, as well as some environmental factors. A survey was conducted at ten elementary schools where the expert system's, psychologists' and teachers' estimates were obtained for each child in the sample. The paper discusses some differences among the estimations of children's mathematical gift obtained by the ES, psychologists and teachers. The results show that the expert system can be suggested as a methodological tool to assist teachers in making decision about children's gift in mathematics., Znanstveno dokazana psihološka detekcija darovitih učenika obično nije dostupna u svim školama. Kako bi se omogućilo točno i rano prepoznavanje matematički darovite djece, predložen je inteligentni ekspertni sustav MathGift kao pomoć učiteljima u donošenju odluke o matematičkoj darovitosti djece u četvrtom razredu osnovne škole. Osim matematičkih kompetencija, sustav uključuje ostale komponente darovitosti u donošenju odluke, kao što su kognitivne komponente darovitosti, osobne komponente koje pridonose razvoju darovitosti, strategije učenja i vježbanja, kao i neke činitelje okoline. Istraživanje je provedeno u deset osnovnih škola, pri čemu su prikupljene procjene ekspertnoga sustava, psihologa i učitelja za svako dijete u promatranom uzorku. Rad opisuje neke razlike između procjena darovitosti djece dobivenih u ekspertnom sustavu psihologa i učitelja. Rezultati pokazuju da se ekspertni sustav može predložiti kao metodološki alat koji će pomoći učiteljima u odlučivanju o matematičkoj darovitosti djece., Eine objektiv nachweisbare psychologische Detektierung begabter Schüler ist an Schulen gemeinhin nicht möglich. Um eine exakte Früherkennung mathematisch begabter Kinder zu ermöglichen, wurde vorgeschlagen, Grundschullehrern der vierten Klasse das Expertensystem MathGift als Hilfsmittel an die Hand zu geben, um ihnen den Einblick in die Begabung von Schülern zu erleichtern. Neben mathematischen umfasst das Expertensystem auch andere zur Begabungsentfaltung beitragende Kompetenzen, so kognitive Fähigkeiten, persönliche Merkmale, Lern- und Übungsstrategien sowie einige Faktoren aus dem sozialen Umfeld. Eine entsprechende Untersuchung wurde an zehn kroatischen Grundschulen durchgeführt, wobei die auf ein bestimmtes Kind bezogenen Resultate des Expertensystems sowie Einschätzungen von Psychologen und Lehrern festgehalten wurden. Die Verfasser beschreiben, welche Abweichungen dabei zu verzeichnen waren. Die Untersuchungsergebnisse zeigen, dass das Expertensystem als methodologisches Hilfsmittel zur Erkennung mathematischer Begabung vorgeschlagen werden kann.

The major challenge in influenza vaccination is to predict vaccine efficacy. The purpose of this study was to design a model to enable successful prediction of the outcome of influenza vaccination based on real historical medical data. A non-linear neural network approach was used, and its performance compared to logistic regression. The three neural network algorithms were tested: multilayer perceptron, radial basis and probabilistic in conjunction with parameter optimization and regularization techniques in order to create an influenza vaccination model that could be used for prediction purposes in the medical practice of primary health care physicians, where the vaccine is usually dispensed. The selection of input variables was based on a model of the vaccine strain which has frequently been changed and on which a poor influenza vaccine response is expected. The performance of models was measured by the average hit rate of negative and positive vaccine outcome. In order to test the generalization ability of the models, a 10-fold cross-validation procedure revealed that the model obtained by multilayer perceptron produced the highest average hit rate among neural network algorithms, and also outperformed the logistic regression model with regard to sensitivity and specificity. Sensitivity analysis was performed on the best model and the importance of input variables was discussed. Further research should focus on improving the performance of the model by combining neural networks with other intelligent methods in this field.

The paper deals with the problem of predicting the time to default in credit behavioural scoring. This area opens a possibility of including a dynamic component in behavioural scoring modelling which enables making decisions related to limit, collection and recovery strategies, retention and attrition, as well as providing an insight into the profitability, pricing or term structure of the loan. In this paper we compare survival analysis and neural networks in terms of modelling and results. The neural network architecture is designed such that its output is comparable to the survival analysis output. Six neural network models were created, one for each period of default. A radial basis neural network algorithm was used to test all six models. The survival model used a Cox modelling procedure. Further, different performance measures of all models were discussed since even in highly accurate scoring models, misclassification patterns appear. A systematic comparison ‘ 3+2+2’ procedure is suggested to find the most effective model for a bank. Additionally, the survival analysis model is compared to neural network models according to the relative importance of different variables in predicting the time to default. Although different models can have very similar performance measures they may consist of different variables. The dataset used for the research was collected from a Croatian bank and credit customers were observed during a twelve-month period. The paper emphasizes the importance of conducting a detailed comparison procedure while selecting the best model that satisfies the users' interest.

The paper investigates the possibility of integrating expert systems and neural networks in an intelligent decision support tool that will be able to assist teachers in recognizing mathematically gifted children in elementary schools. The knowledge base of the expert system consisted of five logic blocks describing basic components of a childpsilas mathematical gift identified in authorspsila previous research. The neural network model was created to learn psychological evaluations of children. The performance of individual models was compared using a 10-fold cross-validation procedure. Three neural network algorithms were tested. The results showed that the average hit rate of the expert system was higher than the average hit rate of the neural network model. In order to improve the accuracy of identifying gifted children, a postprocessing procedure was suggested that combines the results of both models. The integrated model was found to be more successful in recognizing mathematically gifted children.

The paper aims to suggest models for predicting computer and Web attitudes using artificial neural networks with an additional support of standard statistics in data preparatiom and future selection. Based on previously confirmed statistical instruments CAS and WAS, three models were designed. The first two models observed computer and Web attitudes separately, while both attitudes were observed in the third model. Demographic characteristics, experiences with different computer skills and Internet were included as input variables. The results show that neural network models are able to predict computer and Web attitudes more accurately than linear models. The selection of features confirmed some previous, but also revealed some new predictors tested on a Croatian dataset.

After production and operations, finance and investments are one of the most frequent areas of neural network applications in business. The lack of standardized paradigms that can determine the efficiency of certain NN architectures in a particular problem domain is still present. The selection of NN architecture needs to take into consideration the type of the problem, the nature of the data in the model, as well as some strategies based on result comparison. The paper describes previous research in that area and suggests a forward strategy for selecting best NN algorithm and structure. Since the strategy includes both parameter-based and variable-based testings, it can be used for selecting NN architectures as well as for extracting models. The backpropagation, radial-basis, modular, LVQ and probabilistic neural network algorithms were used on two independent sets: stock market and credit scoring data. The results show that neural networks give better accuracy comparing to multiple regression and logistic regression models. Since it is model-independant, the strategy can be used by researchers and professionals in other areas of application.

The aim of this paper is to use a questionnaire to gather subjective information regarding health promotion and disease prevention in the Republic of Croatia. The research was conducted on a sample of 980 participants from the entire Republic of Croatia. A written questionnaire was used for the purpose of gath-ering information. Descriptive and inferential statistical methods were used for data analysis. Descriptive statistical indicators were determined for all research variables, based on participants’ responses, which were measured on a 5-point scale. Differences considered statistically significant were confirmed on the sig-nificance level of p < 0.05. Participants largely agree that they need additional information regarding health promotion and disease prevention. Approximately 81% of participants expect to be informed at family physician offices about activities related to health promotion and disease prevention, while almost 84% of participants mostly or completely agree that they are willing to receive magazines issued by a health institution. Furthermore, there is also a significant number of partici-pants who would respond to an invitation by a health institution into Health promotion counseling or disease prevention counseling and participate in pub-lic discussions on health promotion and disease prevention. Approximately 57% of participants agree with the statement that there is a form in which they can write their compliments, suggestion, complaints or remarks regarding health promotion and disease prevention, while slightly more than a fifth of participants fully or mostly disagree with that statement. The concept of health assumes a completely new dimension with citizen active participation. It emphasizes the positive potential abilities of individual and community too, via active creation of health policy, state their needs regarding health promotion, on whose path the health is quarantined as a basic human right.

In the 21st century, the century of megatrends, tourism has been going through a certain structural transformation. Almost every day, the sector of tourism faces certain changes that leave deep and strong implications, both positive and negative. Although mass tourism still remains the leading form of tourism in the global tourism market, it is – so to say – being increasingly opposed to by specific forms of tourism that are based on the platform of meeting the tourist’s needs and experiences. However, mass tourism will never be fully replaced by specific forms of tourism – they will only stand as strong alternatives to mass tourism. One such form of tourism is certainly agritourism – the most rapidly growing form of rural tourism worldwide. Precisely agritourism has been of a great interest to a growing number of international scientists in the last three decades. In the Republic of Croatia and its academic community, this form of tourism has unjustifiably not gained the place and the significance that it deserves. There is namely a very scarce base of theoretical, as well as empiric research. Therefore, the main goal of this work is to represent the present knowledge and empirical results of Croatian scientists and experts, which were collected in various Croatian tourist regions, at the theoretical and research level. The collected knowledge will be analyzed by use of corresponding scientific methods.

Widespread mobile and web technology is used frequently to (i) present learning content, (ii) facilitate communication / interaction between learners and educators and (iii) assess learning outcomes, regardless of the topic or the context of learning. Data generated in the process is commonly stored and available for subsequent analysis. Increasing volumes of data about learning and teaching processes generated in different educational contexts (whether formal or informal, higher education or lifelong learning) led to advent of concepts such as learning analytics (LA) and educational data mining (EDM). LA and EDM develop rapidly due to advances in data management, especially in terms of managing large data sets since it enables storing and organizing educational data in an easy way. In addition to modern day tools for data management, a large number of tools for (big) data analysis are available on the market. In higher education, one of the key questions is how to increase student engagement and, in the process, achieve transformative learning outcomes. LA and EDM are crucial tools for answering such a complex question. Approaches behind LA and EDM are very similar. The key difference is that LA relies on human interpretation of the data, data visualization and social network analysis, whereas EDM is based on automated machine learning / data mining methods. Drawing on number of success stories and research reports, we explore and assess appropriateness of LA and EDM concepts and tools in one higher education institution (HEI) in Croatia. Analysis that goes beyond basic reports provided within centrally managed academic information systems is advocated and potential benefits are investigated and explained. As a proof of concept, a data mart that combines data from couple of data sources (Moodle LMS and custom academic information system) is devised and presented as a basis for systemic, real-time analysis of educational data in HEIs.