| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

Search the digital library catalog Help

Query: search in
search in
search in
search in
* old and bologna study programme


1 - 10 / 19
First pagePrevious page12Next pageLast page
Adaptive boosting method for mitigating ethnicity and age group unfairness
Ivona Colakovic, Sašo Karakatič, 2024, original scientific article

Abstract: Machine learning algorithms make decisions in various fields, thus influencing people’s lives. However, despite their good quality, they can be unfair to certain demographic groups, perpetuating socially induced biases. Therefore, this paper deals with a common unfairness problem, unequal quality of service, that appears in classification when age and ethnicity groups are used. To tackle this issue, we propose an adaptive boosting algorithm that aims to mitigate the existing unfairness in data. The proposed method is based on the AdaBoost algorithm but incorporates fairness in the calculation of the instance’s weight with the goal of making the prediction as good as possible for all ages and ethnicities. The results show that the proposed method increases the fairness of age and ethnicity groups while maintaining good overall quality compared to traditional classification algorithms. The proposed method achieves the best accuracy in almost every sensitive feature group. Based on the extensive analysis of the results, we found that when it comes to ethnicity, interestingly, White people are likely to be incorrectly classified as not being heroin users, whereas other groups are likely to be incorrectly classified as heroin users.
Keywords: fairness, boosting, machine learning, classification
Published in DKUM: 24.05.2024; Views: 103; Downloads: 5
.pdf Full text (1,66 MB)
This document has many files! More...

Reduction of Neural Machine Translation Failures by Incorporating Statistical Machine Translation
Jani Dugonik, Mirjam Sepesy Maučec, Domen Verber, Janez Brest, 2023, original scientific article

Abstract: This paper proposes a hybrid machine translation (HMT) system that improves the quality of neural machine translation (NMT) by incorporating statistical machine translation (SMT). Therefore, two NMT systems and two SMT systems were built for the Slovenian-English language pair, each for translation in one direction. We used a multilingual language model to embed the source sentence and translations into the same vector space. From each vector, we extracted features based on the distances and similarities calculated between the source sentence and the NMT translation, and between the source sentence and the SMT translation. To select the best possible translation, we used several well-known classifiers to predict which translation system generated a better translation of the source sentence. The proposed method of combining SMT and NMT in the hybrid system is novel. Our framework is language-independent and can be applied to other languages supported by the multilingual language model. Our experiment involved empirical applications. We compared the performance of the classifiers, and the results demonstrate that our proposed HMT system achieved notable improvements in the BLEU score, with an increase of 1.5 points and 10.9 points for both translation directions, respectively.
Keywords: neural machine translation, statistical machine translation, sentence embedding, similarity, classification, hybrid machine translation
Published in DKUM: 20.02.2024; Views: 220; Downloads: 21
.pdf Full text (400,40 KB)
This document has many files! More...

Acoustic Gender and Age Classification as an Aid to Human–Computer Interaction in a Smart Home Environment
Damjan Vlaj, Andrej Žgank, 2023, original scientific article

Abstract: The advanced smart home environment presents an important trend for the future of human wellbeing. One of the prerequisites for applying its rich functionality is the ability to differentiate between various user categories, such as gender, age, speakers, etc. We propose a model for an efficient acoustic gender and age classification system for human–computer interaction in a smart home. The objective was to improve acoustic classification without using high-complexity feature extraction. This was realized with pitch as an additional feature, combined with additional acoustic modeling approaches. In the first step, the classification is based on Gaussian mixture models. In thesecond step, two new procedures are introduced for gender and age classification. The first is based on the count of the frames with the speaker’s pitch values, and the second is based on the sum of the frames with pitch values belonging to a certain speaker. Since both procedures are based on pitch values, we have proposed a new, effective algorithm for pitch value calculation. In order to improve gender and age classification, we also incorporated speech segmentation with the proposed voice activity detection algorithm. We also propose a procedure that enables the quick adaptation of the classification algorithm to frequent smart home users. The proposed classification model with pitch values has improved the results in comparison with the baseline system.
Keywords: acoustic classification, acoustic signal processing, Gaussian mixture model, pitch analysis, smart home
Published in DKUM: 11.12.2023; Views: 359; Downloads: 18
.pdf Full text (2,07 MB)
This document has many files! More...

Towards the classification of self-sovereign identity properties
Špela Čučko, Šeila Bećirović, Aida Kamišalić Latifić, Saša Mrdović, Muhamed Turkanović, 2022, original scientific article

Abstract: Self-Sovereign Identity (SSI) is a novel and emerging, decentralized digital identity approach that enables entities to control and manage their digital identifiers and associated identity data while enhancing trust, privacy, security, and the many other properties identified and analyzed in this paper. The paper provides an overview and classification of the SSI properties, focusing on an in-depth analysis, furthermore, presenting a comprehensive collection of SSI properties that are important for the implementation of the SSI system. In addition, it explores the general SSI process flow, and highlights the steps in which individual properties are important. After the initial purification and classification phase, we then validated properties among experts in the field of Decentralized and Self-Sovereign Identity Management using an online questionnaire, which resulted in a final set of classified and verified SSI properties. The results can be used for further work on definition and standardization of the SSI field.
Keywords: classification, credential, decentralized, identity, identified, principles, properties, selfsovereign, verifiable
Published in DKUM: 22.09.2023; Views: 214; Downloads: 32
.pdf Full text (5,62 MB)
This document has many files! More...

Scoping review on the multimodal classification of depression and experimental study on existing multimodal models
Umut Arioz, Urška Smrke, Nejc Plohl, Izidor Mlakar, 2022, review article

Abstract: Depression is a prevalent comorbidity in patients with severe physical disorders, such as cancer, stroke, and coronary diseases. Although it can significantly impact the course of the primary disease, the signs of depression are often underestimated and overlooked. The aim of this paper was to review algorithms for the automatic, uniform, and multimodal classification of signs of depression from human conversations and to evaluate their accuracy. For the scoping review, the PRISMA guidelines for scoping reviews were followed. In the scoping review, the search yielded 1095 papers, out of which 20 papers (8.26%) included more than two modalities, and 3 of those papers provided codes. Within the scope of this review, supported vector machine (SVM), random forest (RF), and long short-term memory network (LSTM; with gated and non-gated recurrent units) models, as well as different combinations of features, were identified as the most widely researched techniques. We tested the models using the DAIC-WOZ dataset (original training dataset) and using the SymptomMedia dataset to further assess their reliability and dependency on the nature of the training datasets. The best performance was obtained by the LSTM with gated recurrent units (F1-score of 0.64 for the DAIC-WOZ dataset). However, with a drop to an F1-score of 0.56 for the SymptomMedia dataset, the method also appears to be the most data-dependent.
Keywords: multimodal depression classification, scoping review, real-world data, mental health
Published in DKUM: 11.08.2023; Views: 429; Downloads: 44
.pdf Full text (1,43 MB)
This document has many files! More...

Improved Boosted Classification to Mitigate the Ethnicity and Age Group Unfairness
Ivona Colakovic, Sašo Karakatič, 2022, published scientific conference contribution

Abstract: This paper deals with the group fairness issue that arises when classifying data, which contains socially induced biases for age and ethnicity. To tackle the unfair focus on certain age and ethnicity groups, we propose an adaptive boosting method that balances the fair treatment of all groups. The proposed approach builds upon the AdaBoost method but supplements it with the factor of fairness between the sensitive groups. The results show that the proposed method focuses more on the age and ethnicity groups, given less focus with traditional classification techniques. Thus the resulting classification model is more balanced, treating all of the sensitive groups more equally without sacrificing the overall quality of the classification.
Keywords: fairness, classification, boosting, machine learning
Published in DKUM: 02.08.2023; Views: 402; Downloads: 29
.pdf Full text (884,95 KB)
This document has many files! More...

K-vertex: a novel model for the cardinality constraints enforcement in graph databases : doctoral dissertation
Martina Šestak, 2022, doctoral dissertation

Abstract: The increasing number of network-shaped domains calls for the use of graph database technology, where there are continuous efforts to develop mechanisms to address domain challenges. Relationships as 'first-class citizens' in graph databases can play an important role in studying the structural and behavioural characteristics of the domain. In this dissertation, we focus on studying the cardinality constraints mechanism, which also exploits the edges of the underlying property graph. The results of our literature review indicate an obvious research gap when it comes to concepts and approaches for specifying and representing complex cardinality constraints for graph databases validated in practice. To address this gap, we present a novel and comprehensive approach called the k-vertex cardinality constraints model for enforcing higher-order cardinality constraints rules on edges, which capture domain-related business rules of varying complexity. In our formal k-vertex cardinality constraint concept definition, we go beyond simple patterns formed between two nodes and employ more complex structures such as hypernodes, which consist of nodes connected by edges. We formally introduce the concept of k-vertex cardinality constraints and their properties as well as the property graph-based model used for their representation. Our k-vertex model includes the k-vertex cardinality constraint specification by following a pre-defined syntax followed by a visual representation through a property graph-based data model and a set of algorithms for the implementation of basic operations relevant for working with k-vertex cardinality constraints. In the practical part of the dissertation, we evaluate the applicability of the k-vertex model on use cases by carrying two separate case studies where we present how the model can be implemented on fraud detection and data classification use cases. We build a set of relevant k-vertex cardinality constraints based on real data and explain how each step of our approach is to be done. The results obtained from the case studies prove that the k-vertex model is entirely suitable to represent complex business rules as cardinality constraints and can be used to enforce these cardinality constraints in real-world business scenarios. Next, we analyze the performance efficiency of our model on inserting new edges into graph databases with varying number of edges and outgoing node degree and compare it against the case when there is no cardinality constraints checking. The results of the statistical analysis confirm a stable performance of the k-vertex model on varying datasets when compared against a case with no cardinality constraints checking. The k-vertex model shows no significant performance effect on property graphs with varying complexity and it is able to serve as a cardinality constraints enforcement mechanism without large effects on the database performance.
Keywords: Graph database, K-vertex cardinality constraint, Cardinality, Business rule, Property graph data model, Property graph schema, Hypernode, Performance analysis, Fraud detection, Data classification
Published in DKUM: 10.08.2022; Views: 647; Downloads: 76
.pdf Full text (3,43 MB)

Formal boundaries of Slovenian law
Bojan Tičar, 2018, review article

Abstract: Purpose: The article is primarily intended for foreign exchange students (e.g. students participating in the Erasmus+ programme) at the Faculty of Criminal Justice and Security of the University of Maribor studying how Slovenia regulates the field of criminal justice and security. The article deals with the formal boundaries of Slovenian law mostly from the viewpoint of the legal order in force in the country. Readers will learn how the Slovenian legal order functions, which general acts are adopted by the state and which by local bodies, the rules governing their application, and the relationships between them, as well as the way EU law is applied in Slovenia. Design/Methods/Approach: The article is a review article based on a descriptive analytical method and linguistic interpretation of the relevant regulations. The author also applies a historical method – primarily by presenting the Roman law perspective on legal concepts – as well as teleological and legal philosophical methods in defining legal concepts. Findings: The article examines fundamental legal institutions. The author establishes the attitude of writers in the fields of the theory of law and critical jurisprudence regarding the definitions of key legal concepts and phenomena. The article concludes with Kant’s remark that lawyers are still seeking a definition of their concept of law (Perenič, 2007). Research Limitations / Implications: The article is short. The legal definitions are occasionally simplified which, however, is not to the reader’s disadvantage. In some instances, the author attempts to simplify complicated legal concepts with the objective of making the study of other subjects in the field of criminal justice and security easier for the reader and to provide a clear foundation for a basic understanding of the categorical apparatus in law. Practical Implications: The article has practical value for foreign, English-speaking students who, generally, are not students of law, but need a basic understanding of the fundamental legal concepts useful in most social science research. The definitions of the concepts are appropriate and contemporary and thereby contribute to better understanding of the field. Originality/Value: The article is a review article and therefore its originality is limited. The author namely does not establish any new scientific findings, but summarises and defines already known concepts. The article’s original value is that the author presents fundamental legal concepts and definitions in a readable and easy-to-digest manner that the reader can easily remember. The definitions of the legal concepts addressed in the article are precise and useful and will serve the reader in further study or research.
Keywords: law, legal order, morality, classification of law, sources of law, legal norm, statute, lex specialis
Published in DKUM: 20.04.2020; Views: 1155; Downloads: 66
.pdf Full text (337,32 KB)
This document has many files! More...

Systems approach to standardisation, classification and modelling of managed events for tourism
Tadeja Jere Lazanski, Aleksandra Golob, 2015, original scientific article

Abstract: Background and Purpose: The standardisation and classification of managed events provide a legislative basis to distinguish events managed for tourism in their characteristics and quality. The systems approach to standardisation and classification of managed events is a unique, holistic view of event management quality and event organization in tourism. It enables a clear overview of a researched topic and provides adequate support to design and decision-making. In this paper, we explain the meaning of standardisation and classification for Slovenian legislation related to event management. We present the importance of a systems approach methodology for event categorization and classification as it relates to the quality of event management organization, the quality of staff, the quality of the event program and the quality of event services. Objectives: Provide an overview of events in tourism, related definitions and information gathered from scientific authors, which serves as current systems approach principles with which we want to achieve the desired results, positive changes in legislation; in our case-in the field of managed event quality for tourism through standardisation and classification of events on the national level in Slovenia. Method: A descriptive method and systems approach methods are fundamental methodological principles in our analysis. In the context of a systems approach, we used qualitative modelling and constructed causal loop models (CLD) of the legislative system of events and investments in the events. We also used context-dependent modelling (SD model) in a frame of systems dynamics. Results: We present the most appropriate solution to eliminate our problem or question about how to achieve high quality and unique events within event tourism and with event management, thereby creating added value to an event legislative system. We explain suggestions for achieving triple-bottom elements through well-designed quality standards and classification of events, which leads to an optimal categorization of events. Conclusion: From a systems point of view, event tourism processes, including event management, are systems consisting of people and technologies with the purpose of designing, producing, trading and deploying the idea of an event. It is necessary to transform the current Slovenian legislative system of events and prepare a document which standardizes and classifies events based on systems approach methodology.
Keywords: systems approach, standarization, classification, tourism, managed events, modelling
Published in DKUM: 22.01.2018; Views: 1313; Downloads: 364
.pdf Full text (473,40 KB)
This document has many files! More...

Macroeconomic determinants of the non-performing placements and off-balance sheet liabilities of Croatian banks
Manuel Benazić, Dajana Radin, 2015, original scientific article

Abstract: Background and Purpose: The non-performing placements and off-balance sheet liabilities are often considered key factors that lead to banking crises. Economic and financial crises increase the level of the non-performing placements and off-balance sheet liabilities which can cause significant losses for banks. Effective management and regulatory/ supervisory institutions such central banks should be able to recognize and quantify these effects. Therefore, the purpose of this study is to empirically determine the existence and the quantitative impact of main Croatian macroeconomic variables on the non-performing placements and off-balance sheet liabilities of Croatian banks in the long and short-run. Methodology: For this purpose the bounds testing (ARDL) approach for cointegration is applied. The ARDL model is performed in two steps. The first step starts with conducting the bounds test for cointegration. In the second step, when cointegration is found, the long-run relationship and the associated error correction model are estimated. Results: The results indicate the existence of stable cointegration relationship between the variables i.e. in the longrun, an increase in real GDP reduces the level of the non-performing placements and off-balance sheet liabilities of Croatian banks wherein an increase in prices, unemployment, interest rate and the depreciation of the Croatian kuna exchange rate increases their level. On the other hand, in the short-run the results are rather mixed. Conclusion: To avoid crises, effective bank management and regulatory/supervisory institutions should be able to recognize and quantify these effects. This is a necessary precondition for implementation of an adequate prudential and monetary policy measures for reducing the level of the non-performing placements and off-balance sheet liabilities.
Keywords: non-performing placements and off-balance sheet liabilities, non-performing loans, economic and financial crises, credit risk, classification of placements and off-balance sheet liabilities
Published in DKUM: 04.12.2017; Views: 1137; Downloads: 356
.pdf Full text (467,97 KB)
This document has many files! More...

Search done in 0.19 sec.
Back to top
Logos of partners University of Maribor University of Ljubljana University of Primorska University of Nova Gorica