Institutional Repository of Academic Research University of Macedonia
Not a member yet
2215 research outputs found
Sort by
A Bayesian approach for the determinants of bitcoin returns
No full textThe aim of this paper is to identify potential determinants of bitcoin returns. We consider a wide range of various determinants including economic, financial and technology-related factors as well as uncertainty and attention indices. The analysis is conducted using LASSO models estimated using both frequentist and Bayesian methods. We evaluate the ability of these estimators to forecast bitcoin returns. The results indicate that a Bayesian LASSO model that takes into account the stochastic volatility and the leverage effect provides the most accurate forecasts. Using this model we are able to identify alternative drivers of bitcoin returns and analyse the underlying mechanisms that affect bitcoin returns.9110303
Enhancing Fraud Detection via GNNs with Synthetic Fraud Node Generation and Integrated Structural Features
No full textGraph Neural Networks are widely employed for node classification in attributed networks. When it comes to fraud detection, however, GNNs can perform poorly, because a node’s features are typically computed based on its local neighborhood, and this allows fraudsters to “blend in” among legitimate users. In this paper, GNNs and supervised contrastive learning are proposed for fraud detection on datasets where fraudsters may use intricate strategies to camouflage themselves within the network. We train our GNNs using novel structural features in addition to those typically used in similar studies. The proposed features are based on the empirical probability distributions of various graph structural attributes which are extracted from a given dataset. We also apply supervised contrastive learning, enhanced with synthetic samples for the minority class (i.e., the fraudsters). Under our approach, the classifying capability of the GNN (measured via F1-macro, AUC, Recall) is improved by boosting the representation power of the calculated embeddings that maximize the similarity between legitimate users while minimizing that between fraudsters and legitimate users. Numerical experiments on two real-world multi-relation graph datasets (Amazon and YelpChi) demonstrate the effectiveness of the proposed method, whose improvements over the state-of the-art were especially significant in the larger YelpChi dataset.15020110125Lecture Notes in Computer Scienc
A note on the determinants of non-fungible tokens returns
No full textWe aim to identify the determinants of non-fungible tokens non-fungible tokens (NFTs) returns. The 10 most popular NFTs based on their price, trading volume, and market capitalisation are examined. Twenty-three potential drivers of the returns of each NFT are considered. We employ a Bayesian LASSO model which takes into account stochastic volatility and leverage effect. The results indicate that NFTs returns are primarily driven by volatility and ethereum returns. We find a weak connection between NFTs returns and conventional assets, such as stock, oil, and gold markets
Unbounded heteroscedasticity in autoregressive models
No full textThis paper develops the asymptotic theory for stable autoregressive models in which the noise variance grows in a polynomial-like fashion. It is shown that the asymptotic distribution of the OLS estimator of the coefficient vector is multivariate normal with a covariance matrix that depends on the order, k, of the variance growth. A consistent estimator of k is proposed, which delivers heteroscedasticity-robust test statistics. The case of “variance decline” is studied as well. It is demonstrated that by means of a simple data transformation producing the time reversed image of the original series, the problem of “variance decrease” can be reformulated in terms of that of polynomial-like variance growth. Simulation evidence suggests that the new procedures work quite well in small samples. Finally, the new methods are used in order to measure potential asymmetries in business cycles dynamics among several OECD countries.29e0035
Creation of knowledge graph from free text using Neo4j and Stanza: An application in machine translation from English to Spanish
No full textThis paper studies the creation of a knowledge graph from free text using tools of Natural Language processing and machine learning. The paper focuses on different aspects of NLP and knowledge graph creation, giving a detailed description of the methodologies and tools used. We present key concepts of NLP such as NLU (Natural Language Understanding) and NLG (Natural language generation) as well as the different tools used such as Stanford NLP (Stanza). We present a practical example of creating a knowledge graph with Neo4j, from a 5 sentences text, and we go through all our used techniques such as the design of the entities and relationships, coding classes and algorithms etc. We use the example text of 5 sentences to create the knowledge graph, and also to propose an modern application of translation our 5 sentences text from English to Spanish.464470Proceedings of the 28th Pan-Hellenic Conference on Progress in Computing and Informatic
Unveiling the power of word-of-mouth in pre-recruitment employer branding strategy during COVID-19
No full textPurpose
This study, based on signaling theory, examines the pre-recruitment employer branding strategy during the COVID-19 pandemic. It investigates the relationship between spontaneous word-of-mouth (WOM) recommendations for companies and prospective candidates' job application intentions. Specifically, the study explores serial mechanisms mediating the characteristics of company online career pages, including the perceived informativeness of online job advertisements (ads), candidates' preferences for its web approach to them and the company’s reputation.
Design/methodology/approach
Reflecting prospective candidates from students and young alumni of universities, partial least squares structural equation modeling (PLS-SEM) was employed on a sample of 737 individuals representing various fields of study from Greek universities.
Findings
The findings highlight the effectiveness of positive WOM recommendations during the initial stages of recruitment, particularly amidst COVID-19 challenges in the labor market, notably impacting young candidates. The study suggests that spontaneous WOM, originating from trustful sources, motivates job seekers to actively engage with the company’s web career channels, seeking information and favorable indications of the company’s approach toward its candidates. Positive WOM, combined with informative content and a friendly communication style, plays a critical role in shaping the company’s reputation. Consequently, this encouragement motivates individuals to start their job search efforts and consider applying for positions within the specific organization.
Practical implications
This research provides valuable empirical evidence in the pre-recruitment field, particularly in unforeseen crisis circumstances such as the COVID-19 pandemic. It examines how spontaneous, positive WOM from sources, like peers and alumni, significantly influences young job seekers' perceptions and preferences regarding the company’s career web channels as sources of information and signals about working conditions. The combination of positive WOM with informative content and a friendly communication style in the web approach plays a crucial role in shaping a positive company reputation. Consequently, this encourages candidates to consider applying for positions within the company.
Originality/value
This research contributes to pre-recruitment studies, especially amidst crises like COVID-19. It examines how positive WOM from trusted sources like peers and alma mater alumni influences young job seekers' views on the company’s career web channels. By emphasizing the importance of combining positive WOM with informative web content and a friendly communication style, the study offers insights into effective recruitment strategies. It highlights the significance of positive and spontaneous WOM in attracting young talent and its impact on job seekers' decision-making, even in uncertain conditions. Overall, it advances recruitment practices for attracting candidates.Η μελέτη εξετάζει, μέσα από το πλαίσιο της signaling theory, τη στρατηγική προ-προσέλκυσης employer branding κατά την περίοδο της πανδημίας COVID-19, εστιάζοντας στον ρόλο του αυθόρμητου θετικού Word-of-Mouth (WOM) στις προθέσεις υποβολής αίτησης υποψηφίων. Με δείγμα 737 φοιτητών και νέων αποφοίτων ελληνικών πανεπιστημίων και αξιοποίηση PLS-SEM, τα ευρήματα δείχνουν ότι το θετικό WOM από αξιόπιστες πηγές ενθαρρύνει τους νέους υποψηφίους να αλληλεπιδρούν με τις διαδικτυακές σελίδες καριέρας των εταιρειών και να αναζητούν ενημερωμένο, φιλικό και αξιόπιστο περιεχόμενο. Ο συνδυασμός θετικών συστάσεων, επαρκούς πληροφόρησης και προσιτής διαδικτυακής επικοινωνίας ενισχύει τη φήμη της εταιρείας και αυξάνει την πιθανότητα οι υποψήφιοι να ξεκινήσουν τη διαδικασία αναζήτησης εργασίας και να εξετάσουν σοβαρά την αίτηση σε συγκεκριμένο οργανισμό.46483384
Local and Global Explainability for Technical Debt Identification
No full textIn recent years, we have witnessed an important increase in research focusing on how machine learning (ML) techniques can be used for software quality assessment and improvement. However, the derived methodologies and tools lack transparency, due to the black-box nature of the employed machine learning models, leading to decreased trust in their results. To address this shortcoming, in this paper we extend the state-of-the-art and-practice by building explainable AI models on top of machine learning ones, to interpret the factors (i.e. software metrics) that constitute a module as in risk of having high technical debt (HIGH TD), to obtain thresholds for metric scores that are alerting for poor maintainability, and finally, we dig further to achieve local interpretation that explains the specific problems of each module, pinpointing to specific opportunities for improvement during TD management. To achieve this goal, we have developed project-specific classifiers (characterizing modules as HIGH and NOT-HIGH TD) for 21 open-source projects, and we explain their rationale using the SHapley Additive exPlanation (SHAP) analysis. Based on our analysis, complexity, comments ratio, cohesion, nesting of control flow statements, coupling, refactoring activity, and code churn are the most important reasons for characterizing classes as in HIGH TD risk. The analysis is complemented with global and local means of interpretation, such as metric thresholds and case-by-case reasoning for characterizing a class as in-risk of having HIGH TD. The results of the study are compared against the state-of-the-art and are interpreted from the point of view of both researchers and practitioners.5082110212
The asymmetric impact of leisure externalities on economic growth
No full textLeisure generates externalities for the economy as a whole, as individuals generally get some (dis-)utility from their leisure-time. However, the sign and the extent of the effect that these externalities have on a specific worker's productivity and on the productivity of all other factors used in combination with labor (hence on long-term economic growth) may be asymmetric across different economic activities. The objective of this paper is to shed light on the impact that sector-specific leisure-time externalities have on the innovation rate, on the sectorial allocation of (skilled) labor, and eventually on the long-run economic growth rate, without making any prior assumption on their sign and magnitude. In the baseline model the growth rate of per capita income moves together with all types of leisure externalities, whereas the innovation rate moves together with (and depends solely on) the R&D-sector-specific leisure externality. From numerical analyses, we conclude that sector-specific leisure-time externalities provide asymmetric effects on the growth rate of real per capita GDP and on the way skilled labor is allocated across different economic activities. The robustness of these conclusions is analyzed by using various definitions of leisure along with different utility functions (including leisure as an argument).30e0038
Performance of Publicly Available Large Language Models on Internal Medicine Board-style Questions
No full textOngoing research attempts to benchmark large language models (LLM) against physicians’ fund of knowledge by assessing LLM performance on medical examinations. No prior study has assessed LLM performance on internal medicine (IM) board examination questions. Limited data exists on how knowledge supplied to the models, derived from medical texts improves LLM performance. The performance of GPT-3.5, GPT-4.0, LaMDA and Llama 2, with and without additional model input augmentation, was assessed on 240 randomly selected IM board-style questions. Questions were sourced from the Medical Knowledge Self-Assessment Program released by the American College of Physicians with each question serving as part of the LLM prompt. When available, LLMs were accessed both through their application programming interface (API) and their corresponding chatbot. Mode inputs were augmented with Harrison’s Principles of Internal Medicine using the method of Retrieval Augmented Generation. LLM-generated explanations to 25 correctly answered questions were presented in a blinded fashion alongside the MKSAP explanation to an IM board-certified physician tasked with selecting the human generated response. GPT-4.0, accessed either through Bing Chat or its API, scored 77.5–80.7% outperforming GPT-3.5, human respondents, LaMDA and Llama 2 in that order. GPT-4.0 outperformed human MKSAP users on every tested IM subject with its highest and lowest percentile scores in Infectious Disease (80th) and Rheumatology (99.7th), respectively. There is a 3.2–5.3% decrease in performance of both GPT-3.5 and GPT-4.0 when accessing the LLM through its API instead of its online chatbot. There is 4.5–7.5% increase in performance of both GPT-3.5 and GPT-4.0 accessed through their APIs after additional input augmentation. The blinded reviewer correctly identified the human generated MKSAP response in 72% of the 25-question sample set. GPT-4.0 performed best on IM board-style questions outperforming human respondents. Augmenting with domain-specific information improved performance rendering Retrieval Augmented Generation a possible technique for improving accuracy in medical examination LLM responses.39e000060
Capital structure, voluntary corporate governance and credit ratings: evidence from non-listed SMEs
No full textThis study investigates the capital structure of 26,335 Greek non-listed SMEs during the period 2014–2018 employing a gamut of firm-specific, credit ratings and corporate governance variables. Employing both static and dynamic panel data regression models, the results show that the short-term debt ratio is negatively (positively) related with profitability, tangibility and growth (firm size). The long-term ratio is positively (negatively) associated with profitability, tangibility, and firm age (firm size). Board size exerts a positive effect on the long-term debt ratio, while CEO gender is negatively related to the long-term debt ratio. Finally, higher credit ratings are associated with more debt levels.161174