ISD 2025: 33RD INTERNATIONAL CONFERENCE ON INFORMATION SYSTEMS DEVELOPMENT
PROGRAM FOR FRIDAY, SEPTEMBER 5TH
Days:
previous day
all days

View: session overviewtalk overview

08:00-09:00Registration
09:00-10:00 Session 13: Keynote talk 3

Mirjana Ivanović, University of Novi Sad, Serbia

Title: How Artificial Intelligence Changes Medical Diagnosis, Treatment, and Education

Abstract: In modern dynamic, constantly developing society more and more people suffer from noncommunicable diseases, such as cancer, heart disease, neurodegenerative disease, diabetes, and so on. They are the leading cause of death worldwide and represent an emerging global health threat. Moreover, the population in a lot of countries is getting older and needs additional and special medical and health care. Accordingly, world renowned health stakeholders have recognized the importance of development of wide range of systems/architectures/services that can support diagnoses and treatment decisions for patients, help them in everyday activities and improve their quality of life. For the development of smart and intelligent medical architectures and decision support systems collection of huge amounts of patient’s complex big data is necessary. Such data should be properly curated, aggregated, analyzed and results achieved by AI-based processing of such data are used by the doctors to recommend adequate treatment and actions in order to improve patient’s health status. Contemporary Healthcare and Medical systems are predominantly distributed. Together with intensive use of Machine Learning approaches (ML), Federated Learning (FL) plays an essential role in such systems. Each medical institution might have a lot of patient data, but it can be not enough to train their own local prediction models. Accordingly, the combination of FL and prediction of future patients’ status but also patients’ treatment for achieving satisfactory quality of life parameters is good solutions to break down the barriers of analysis throughout different hospitals. On the other hand, modern AI-based approaches are also employed in the development of systems and environments that can have significant impact on medical and healthcare education, like agents, extended reality, metaverse, and so on. In this talk all the above-mentioned aspects will be considered and some personal experiences in developing such systems will be presented.

10:00-11:00 Session 14: Poster Session P3
Classification of SAX-like Time Series Bitmaps with Siamese Convolutional Neural Networks

ABSTRACT. This research paper concerns the classification of multiclass univariate time series. It extends the existing time series imaging method and describes the process of obtaining new transformations of this sort. The presented techniques leverage the well-known Symbolic Aggregate Approximation representation, which transforms time series from numerical to symbolic domain. The obtained symbolic approximations are later turned into images. After raw time series data is transformed into two-dimensional grayscale bitmaps, these bitmaps are used as input for two alternative deep learning classification approaches. Our study focuses on comparing regular Convolutional Neural Networks and Siamese Neural Networks as time series classifiers. Experimental studies and comparative analyses were performed on well-known, publicly available datasets.

Inference processes in rule cluster knowledge base - various approaches
PRESENTER: Igor Gaibei

ABSTRACT. This paper presents a novel approach to optimize inference in rule-based knowledge systems by introducing a clustering mechanism for rule organization. Rules are clustered using K-Means or Agglomerative Hierarchical Clustering (AHC) algorithms, with different distance measures and clustering strategies. We propose and evaluate four inference strategies based on different group representation methods (mean or median) and rule activation strategies (activation of one or all matching rules). Experimental studies on real knowledge bases show that clustering significantly improves inference performance while maintaining a satisfied inference success rate.

Estimating the value of commercial real estate using machine learning approaches

ABSTRACT. The paper addresses the problem of predicting the attractiveness of the commercial real estate market based on market valuation. The aim is to estimate market value, and models of the commercial real estate market need to be established. The complexity of the problem lies in accounting for market changes resulting from the COVID-19 pandemic and, more broadly, its variability in response to geopolitical shifts, which are reflected in the available data. To address this, a machine learning-based framework is proposed. The proposed models for solving the problem have been validated using data from recent years. Finally, a summary is provided.

Enhancing Network Intrusion Detection through Data Dimensionality Reduction Using Classical and Deep Learning Approaches

ABSTRACT. Network intrusion detection systems face high-dimensional traffic, which degrades accuracy and raises computational costs. We evaluate Principal Component Analysis (PCA), Recursive Feature Elimination (RFE), and a deep autoencoder on the UNSW-NB15 dataset using eight classifiers. RFE delivers peak accuracy (92.5%) with minimal variability. PCA restores near-baseline accuracy while preserving 95% variance with minor tuning. The autoencoder yields nonlinear embeddings but demands extensive training and trails classical methods. These findings guide the selection of reduction strategies under accuracy requirements and resource constraints.

Deep Neural Networks for Automatic Detection and Classification of Laryngeal Pathologies in Endoscopic Imaging

ABSTRACT. This study explores the application of artificial intelligence (AI) methods for the automated detection and classification of laryngeal pathologies in fiberoptic laryngoscopy videos. From recordings of 292 patients, a total of 885 informative image frames were automatically ex- tracted, and subsequently segmented manually by experienced clinicians. Seven distinct pathol- ogy categories were examined using two deep learning models, Mask R-CNN, designed for classification, object detection, and segmentation tasks; and EfficientNet V2L, solely for clas- sification. For the classification task, an across-class average imbalance-resistant F1-score was higher for Mask R-CNN model, 0.95 (confidence interval, CI: 0.90–0.98), than for Efficient- Net V2L 0.74 (CI: 0.66-0.81; McNemar’s test p<0.001). In object detection, a mean average precision of 0.36 (CI: 0.35-0.37) was achieved at an intersection over union threshold of 50%. However, segmentation models reached lower performance, average precision 0.29 (0.28-0.30). In sum, for the larynx pathology analysis, DNNs show more potential for classification than segmentation tasks, with an advantage of Mask R-CNN over EfficientNet architecture.

Evaluation of User Experience with RAG-based Chatbots for Searching Documentation: Industrial Case Study
PRESENTER: Marko Vještica

ABSTRACT. Artificial Intelligence (AI) has accelerated digital transformation across industries, with Large Language Models (LLMs) powering content generation, summarization, and dialogue systems, yet struggling with domain-specific knowledge. In industrial settings, Retrieval-Augmented Generation (RAG) architectures, often implemented as chatbots, address this issue by grounding responses in internal company knowledge. Despite increased industrial deployment, user experience evaluations of RAG-based chatbots remain limited, particularly regarding their effectiveness in supporting domain-specific workplace tasks. In this paper, we present a user evaluation of an RAG-based chatbot conducted in a medium-sized company. Employee feedback on chatbot usability and acceptance is analyzed to guide digitalization efforts in future AI-assisted enterprises.

Transfer Learning for Deepfake Detection in Static Facial Images

ABSTRACT. Digital authentication systems that rely on biometric recognition are especially vulnerable to deepfake attacks, which can be used to impersonate legitimate users and bypass security protocols. As deepfake attacks become increasingly sophisticated, detection methods must evolve rapidly. In this paper, we propose the usage of transfer learning instead of standard deep learning to provide a fast response to novel threats. We evaluate 12 approaches, combining three deep neural networks as feature extractors with four traditional machine learning algorithms as classifiers. Finally, the best-performing model, i.e. ConvNeXt with a support vector classifier, is fine-tuned and evaluated on a real-world dataset, demonstrating strong performance.

Vehicle Detection in Parking Lots Using Deep Learning Techniques

ABSTRACT. Today, deep learning methods are being strongly developed and are used for many different tasks. This paper addresses the task of vehicle detection in parking lots. The focus of this study is to evaluate the performance of several versions of the YOLO (You Only Look Once) object detection algorithm on a self-created dataset, using the models in their default configurations with pre-trained weights from the COCO dataset. The dataset contains various lightning and weather conditions such as sunshine, cloudiness, and the presence of snow. Each YOLO version is evaluated using a range of metrics such as precision, recall, F1 score, and FPS. Methods for optimizing the use of models are then proposed and tested. The results demonstrate the trade-offs between detection accuracy and computational efficiency.

Enhancing Solar Magnetogram Retrieval with Deep Semantic Hashing and Hierarchical Graph Indexing

ABSTRACT. We propose a method for content-based retrieval of solar magnetograms using semantic hashing. The approach is based on data from the Helioseismic and Magnetic Imager (HMI) aboard the Solar Dynamics Observatory (SDO), processed using the SunPy and PyTorch libraries. A mathematical representation of the Sun’s magnetic field regions is constructed in the form of a fixed-length vector, enabling efficient similarity comparisons without the need to analyze full-disk images directly. To reduce retrieval time and dimensionality, a fully connected autoencoder is employed to compress a 400-dimensional descriptor into a compact 50-dimensional semantic hash. Experimental results demonstrate the effectiveness of the proposed approach, achieving the highest precision among evaluated state-of-the-art methods. In addition to image retrieval, the method is also applicable to solar image classification tasks.

Interactive Semi-Automatic Labeling of Point Clouds Using Transformer-Based Descriptors

ABSTRACT. This paper presents a descriptor-based method for labeling point clouds using a two-stage transformer architecture. The first stage consists of an encoder that extracts descriptors from point cloud fragments. The second stage, a decoder, assigns labels to these fragments based on both the descriptor of the current fragment and an earlier predefined pattern descriptor. This approach functions as an interactive labeling tool similar to a brush, with the ability to reinforce or weaken the pattern through direct manipulation of its descriptor.

Performance comparison of machine learning algorithms for accurate office real estate price estimation: suggested approaches

ABSTRACT. This paper aims to test various machine learning algorithms on a real-world dataset of office real estate and identify the most accurate one. A comprehensive analysis was conducted using proprietary data on office real estate in Poland, obtained from a leading market intelligence provider specializing in commercial property analytics, covering a 20-year observation period. The research results indicate that among six tested algorithms including Logistic Regression (LR), Linear Discriminant Analysis (LDA), K-Nearest Neighbors (KNN), Decision Tree Classifier (DT), Gaussian Naive Bayes (GNB), Support Vector Machines (SVM), the Decision Tree Classifier (DT) appears to be the best-fit algorithm for selecting factors to estimate office real estate prices.

Toward a DataOps Framework for Enhancing Data Quality in Data Science

ABSTRACT. Data Science (DS) leverages Data Analytics to help organizations extract value from large datasets and enhance performance. A significant focus in DS is on data collection, cleaning, and transformation to create high-quality datasets for analysis. However, traditional manual data preparation methods are often inefficient and error-prone, particularly in Big Data environments. DataOps seeks to automate data lifecycle stages by integrating DevOps practices, enhancing the quality and reliability of data pipelines. This paper proposes a framework for implementing DataOps, demonstrated through a case study on urban mobility analytics.

Photovoltaic Energy Prediction: Evaluating Feature Selection Methods for Enhanced Forecasting

ABSTRACT. Renewable energy, particularly photovoltaic (PV) systems, plays a crucial role in sustainable energy development. Its production is largely dependent on external factors, especially weather conditions, making the forecasting of generated energy a significant research challenge. Selecting appropriate features that influence electricity production can enhance forecasting accuracy. This paper evaluates various feature selection methods relevant to energy output, aiming to identify the most effective selection strategy and determine the most influential variables. Three groups of methods were analyzed: correlation-based statistical methods, ensemble-based importance metrics, and univariate significance tests. The results highlight the importance of choosing suitable feature selection algorithms to improve the accuracy of PV energy production forecasting.

Gender Disparities in Customer Churn Rates: A Rough Neuro-Fuzzy Classifier-based Analysis

ABSTRACT. Customer churn prediction is a critical challenge for businesses aiming to retain valuable customers. This study employs a rough neuro-fuzzy classifier with CA defuzzification to analyze churn behavior, with a particular focus on gender disparities. Thanks to rough set theory, our approach effectively handles incomplete or missing data by utilizing lower and upper approximations, ensuring robust predictions even when feature values are absent. We evaluate feature importance through two distinct methods: directly from the data and via the classifier, to uncover gender-specific patterns in churn behavior. Moreover, we introduce the notion of conditional significance. Our findings reveal notable gender-based differences in the significance of predictive features. Experimental results, validated through ten-fold cross-validation, demonstrate the classifier's ability to manage missing data without imputation, while also underscoring the heightened sensitivity of female customers to feature availability. This research contributes to the growing body of knowledge on gender-driven consumer behavior, offering practical implications for businesses to refine customer relationship management and reduce churn through gender-specific interventions.

Hybrid AI Framework Based on Fuzzy Rough Sets for Two-Dimensional Magnetic Evaluation of Reinforced Concrete Structures Measurements

ABSTRACT. This work presents an intelligent support system for a novel, non-destructive (NDT), 2D method to identify parameters of reinforced concrete (RC) structures. Using association rule analysis (ARA), it detects relationships between signal changes and structure parameter modifications, identifying signal parameters influenced by a single structural parameter. Multitask learning is used to identify concrete cover thickness, reinforcing bar diameter, and steel class. Features are extracted from the three spatial components of magnetic induction via ACO decomposition, which is suited for creating complex databases. Genetic algorithms improve noise resilience in function approximation. Results are shown as Fuzzy Rough Sets. Three vertically placed sensors, combined with AI, enable precise identification of parameters, with changes in one not affecting others.

Optimizing the DFS-based Strategy for Efficient Execution of Counting Queries in Machine Learning Applications

ABSTRACT. Counting queries are fundamental operations widely used in machine learning applications. This paper focuses on optimizing their execution by introducing algorithmic enhancements to the bitmap-based counting query strategy that relies on a Depth-First Search (DFS) traversal. The proposed approach is evaluated through a benchmark involving the execution of random query streams across multiple test datasets. The experimental results demonstrate a significant speedup, with execution times reduced by factors ranging from 1.26× to 2.25×. Furthermore, potential directions for further improving the performance of counting queries on modern high-performance computing (HPC) systems are discussed.

Neural networks based ensemble classifier for phishing link detection

ABSTRACT. This contribution proposes an ensemble classification model which is based on neural networks prediction models and well-known online incremental learning models.The considered neural network models belong to different families, namely long-short term memory, deep feed forward and convolutional neural networks. The incremental learning models considered are Passive Aggressive, Bernoulli Naive Bayes and Stochastic Gradient Descent Classifiers. This paper aims to develop a prediction model that reduces false positives (FP) while maintaining overall model performance. Moreover, the stability of the model over time and its ability to correctly classify phishing links, even if the concept shift occur, are under considerations. The ensemble model shows promising results, demonstrating its superiority over base models. Some proposed models significantly outperform some base models according to statistical tests.

The Importance of Accessible Re-commerce Applications and Websites in Implementing the Circular Economy

ABSTRACT. The aim of the poster is to explore consumer behaviour in the use of modern technologies in the context of the implementation of the principles of a circular economy. The study adopts a hypothesis: the development of a circular economy is supported by the availability of digital e-commerce platforms. The article presents the results of two original surveys conducted between 2020 and 2023 using the CAWI method on nationally representative samples: N= 1115 (2020), 1147 (2023). Respondents were asked min whether they use apps, online auctions and classifieds websites when getting rid of unnecessary products Both in 2020 and 2023, the group of users selling on classifieds websites was around 80%, with online auctions falling in popularity by 11% and using mobile apps by 7%. However, these are the three most popular ways of selling. Traditional ways of selling are also seeing declines in interest and are chosen by far fewer sellers. Traditional sales at a consignment or pawn shop are chosen by around 10% of those surveyed (down 4%), at fairs by 6% (down 2%) and at a stock exchange by around 4% (down 1%).

11:00-11:30Coffee Break
11:30-13:30 Session 15A: T4: Data Science and Machine Learning 6
Location: Room 1
11:30
Assessing the impact of criteria removal on Multi-Criteria Decision-Making stability: A simulation-based sensitivity analysis

ABSTRACT. Selecting a Multi-Criteria Decision-Making (MCDM) method is critical for developing robust Decision Support Systems (DSS), yet limited attention has been given to assessing their stability under structural changes in decision problems. This study proposes a simulation-based framework for evaluating the robustness of MCDM methods when the least important criteria are iteratively removed. Four selected methods, namely Additive Ratio ASsessment (ARAS), COmplex PRoportional ASsessment (COPRAS), Measurement Alternatives and Ranking according to COmpromise Solution (MARCOS), and MultiAttributive Ideal-Real Comparative Analysis (MAIRCA) were tested across thousands of randomized scenarios, with performance assessed through mean ranking correlation, frequency of ranking alterations, and distribution of similarity values. The findings reveal consistent stability trends across methods while identifying differences in sensitivity to criteria reduction. Notably, MAIRCA and COPRAS exhibited more concise performance distributions, suggesting stronger resilience to problem changes. This work addresses a critical gap in understanding method robustness, supporting more informed selection of MCDM techniques for uncertain decision environments and enhancing the reliability of decision-making processes.

11:50
Leveraging GPS Data and Attention-based BiLSTM for Injury Prediction in Professional Football

ABSTRACT. Injuries pose a significant challenge in professional football, affecting player availability, team performance, and club finances. Accurate prediction of injury risk is crucial for implementing effective prevention strategies. This study develops a deep learning model to predict the likelihood of injury in professional football players using data collected through Catapult Sports tracking devices. The research is carried out in collaboration with KKS Lech Poznań, a Polish professional football club. The proposed model architecture combines bidirectional long- and short-term memory (BiLSTM) networks with an attention mechanism to learn from the time series data and predict player injury risk. The model is trained on sequences of Catapult data spanning 14 days before each recorded injury or non-injury event. To address class imbalance, a custom loss function was implemented that balances focal loss and the F_beta score. The model's performance is evaluated on an independent test set, achieving a specificity of 0.90, an accuracy of 0.90, and a recall of 0.40.

12:10
Parking Spot Segmentation Using Deep Learning Techniques

ABSTRACT. Existing computer vision models are developing rapidly and obtaining successful results, but expensive and time-consuming solutions are still used in many cases and applications. Currently, proposed parking management solutions based on computer vision and artificial intelligence are not sufficiently automated and general. They often assume simplifications and specific conditions. This project provides an overview of state-of-the-art in occupancy detection methods and describes the original solution separated into modules. The main advantages are versatility, lack of requirements for the parking structure, high resistance to external conditions, and openness to extension with additional functionalities.

12:30
Variance-Based Analysis of Global Criteria Importance in the ESP-COMET Method

ABSTRACT. Sensitivity analysis is a critical component of Multi-Criterion Decision analysis (MCDA), enabling the evaluation of how variations in input data influence decision outcomes. Although traditional techniques, such as scenario analysis and stability intervals, can enhance robustness, they often fail to capture the full variability inherent in the decision model. To overcome this limitation, we propose a novel framework that applies Sobol sensitivity indices to assess the global importance of criteria in MCDA, with a particular focus on models constructed using the Characteristic Objects Method (COMET). Unlike standard approaches that perturb the decision matrix, our method evaluates the sensitivity of the decision model as a whole. We demonstrate its effectiveness through controlled experiments, including hybrid configurations such as Expected Solution Point (ESP) with COMET, and use first-order (S1) and total effect (ST) Sobol indices. The results show that the variance-based sensitivity analysis provides valuable, complementary insights into the global behavior of the model and the influence of individual criteria. This study introduces a robust analytical tool that enhances the interpretability and methodological depth of decision support systems in MCDA

12:50
Detection of abnormal network flow by the Distributive Aggregations Ensemble Algorithm

ABSTRACT. In today’s digital landscape, Artificial Intelligence (AI) plays a crucial role in addressing cybersecurity challenges faced by IT companies, as the threat of distributed attacks persists despite implementing Network Intrusion Detection Systems (NIDSs). We propose a novel hybrid classifier leveraging distributivity equations to combine k-Nearest Neighbors (kNN), Decision Trees (DT), and Stochastic Gradient Descent (SGD). Evaluated on UNSW-NB15 and SIMARGL2021 datasets, our method demonstrates competitive performance in accuracy, recall, precision, F1-score, and area under ROC curve (AUC) compared to base classifiers and SOTA techniques (Stacking, Soft Voting - Weighted Average Probabilities, Adaptive Boosting (AdaBoost) and Histogram-based Gradient Boosting Classification Tree (HGBC)). Key innovations include a distributivity-based aggregation framework and class-balancing strategy for imbalanced datasets.

13:10
Filling the Gap in Time: Intelligent Imputation of Historical Parish Records
PRESENTER: Adam Kiersztyn

ABSTRACT. The paper presents a method for imputing missing monthly values in historical parish records, based on data from two Prussian towns (Stargard and Słupsk, 1886–1913). The approach com bines 12 simple estimation methods with predictive models, including Random Forest and Gra dient Boosted Trees. The evaluation was performed using synthetic gaps introduced into com plete datasets, with each experiment repeated independently 10 times. The results show that aggregation models significantly reduce the relative error of imputation, with ensemble models achieving average errors below 1%. The solution is general and can be adapted to similar histor ical sources. Potential applications include demographic series reconstruction and the detection of anomalies in archival data.

11:30-13:30 Session 15B: T7: New Topics in IS Development 1
Chair:
Location: Room 2
11:30
Exploring Trust in Explainable AI Systems: The Role of Gender Dynamics and Alibi Choice

ABSTRACT. This study investigates how gender dynamics and alibi choice affect trust in explainable AI systems. In an experiment with 188 participants engaged in a travel group matching task, we manipulated the displayed gender of evaluated individuals and provided some participants with a feigned (alibi) choice over the scope of the system’s explanation. The experiment did not reveal significant differences in trust or its antecedents based on gender dynamics suggesting users’ over-reliance on AI decisions. Additionally, the provision of the alibi choice slightly worsened perceptions of the system, indicating that superficial control options may trigger increased scrutiny rather than enhancing trust.

12:00
Bridging Theory and Practice: Introducing the Heterogeneous Technology Acceptance Model (H-TAM) for Practical Market Research and Product Comparison

ABSTRACT. This paper examines the evolution and limitations of traditional technology acceptance models, notably the Technology Acceptance Model (TAM) and the Unified Theory of Acceptance and Use of Technology (UTAUT). These models have been a valuable frameworks for IS researchers, however their recent theoretical saturation and rigid methodology have constrained their applicability outside of academic discussion. To address this challenge, we propose the Heterogeneous Technology Acceptance Model (H-TAM), a practitioner-oriented framework that leverages established IS theory for real-world market and product research applications. Central to this framework is the novel concept of Impact Scores, a metric that quantifies a product’s relative performance on specific technology acceptance variables against market benchmarks and competitors. Using simulated data, we illustrate how H-TAM captures differences in user acceptance across technology products, generating insights that directly inform practitioner decision-making in market analysis and product development. This framework advances the practical utility of technology acceptance theory and offers a foundation for more context-sensitive research and applications.

12:30
Free/Libre and Open Source Software in the Social Sector: Towards a Maturity Model

ABSTRACT. The advantages of Free/Libre and Open Source Software (FLOSS) in creating accessible, sustainable, and community-oriented technological solutions have captured the attention of the social sector. However, security issues, licensing, training, maintenance, and insufficient use cases require new tools to assist these organizations. This paper identifies which dimensions can be used to develop a FLOSS maturity model for the social sector. Design science research was selected, starting with a systematic literature review on FLOSS maturity models. The study involved four partner organizations from the social sector in Portugal and Angola. Leveraging these findings, we developed a conceptual framework that aligns services usually offered by the social sector with the dimensions that support FLOSS adoption. The evaluation used the system usability scale questionnaire, and the results were further assessed through semi-structured interviews. Our research can contribute to expanding FLOSS adoption in a mission-critical sector struggling to define the best digital transformation strategy.

13:00
A Gamified Application for Test Anxiety in Computer Science Education: A Pilot Study

ABSTRACT. This paper presents YoungMind, a gamified digital application designed to assist Computer Science undergraduate students in practising anxiety-reduction techniques for managing test anxiety, particularly when they face programming tests. A pilot study involving 15 students was conducted to evaluate the application's effectiveness, assess the impact of its game elements, and explore its potential for supporting anxiety management. The study compared two groups: one using the gamified version and the other using a non-gamified version with identical content. Student feedback gathered through a user experience survey indicated that the gamified version was more engaging and facilitated easier learning and practice of anxiety-reduction skills, suggesting that YoungMind may offer benefits in helping students manage test anxiety.

11:30-13:30 Session 15C: T8: Business Models and IS for Smart Environments 1
Location: Room 3
11:30
Energy consumers’ perspectives on social, environmental and quality of service factors - an analysis of linkages in developed and developing countries

ABSTRACT. This paper aims to assess the role of social, quality energy services and environmental factors in household energy decision-making. Authors identified links between these dimensions for energy consumers in making decisions and assessed differences between advanced and developing economies. A questionnaire-based survey using the CAPI technique was used. The survey was conducted among respondents from Bulgaria, England, Jordan, Poland, Russia and the United Arab Emirates. The correlation and Principal Component Analyses were used to analyze the data. Results show significant interdependencies between consumers' propensity to government social support, quality of energy services and environmental aspects. Energy consumers who advocate for increased subsidies for the poorest also support banning disconnections due to non-payment. Additionally, the study reveals that consumers highly value pro-quality tools, mainly online energy management tools and artificial AI-driven solutions. The findings contribute valuable insights for policymakers and energy sector practitioners in addressing consumer needs and designing sustainable energy policies.

12:00
Development of digital business models in the context of digital transformation conditions

ABSTRACT. The aim of this article is to identify the directions of development of digital business models. The article begins with a theoretical overview, highlighting the impact of digital transformation on the evolution of digital business models and presenting their various definitions. Next, a bibliometric analysis was conducted using VOSviewer software, aimed at verifying the thematic areas co-occurring in the literature on the subject along with the concept of a digital business model. The study was based on 163 scientific publications included in the Scopus database. The analysis focused on the co-occurrence of keywords, with attention to such attributes as: number of occurrences and total link strength. Additionally, a network visualization map was presented in graphic form, which allowed for clustering all keywords adequately to the interdependencies and frequency of occurrence between them. The final part of the article presents conclusions, which allowed us to notice that research on digital business models encompasses a wide range of issues—both theoretical, such as digital transformation, and empirical, including the implementation of new technologies.

12:30
Artificial Intelligence in Warehouse Logistics as a Tool for Spatial Structure Optimization

ABSTRACT. The aim of the study was to quantitatively determine the impact of implementing artificial intelligence (AI) solutions on the operational efficiency of logistics warehouses. To achieve this, an extended linear regression model was developed, which, in addition to classical determinants such as the level of automation and service quality, included a variable describing the degree of AI technology integration in the analyzed units. The model was estimated using the Ordinary Least Squares (OLS) method on a synthetic dataset (n = 100), generated based on average statistics from official sources and market proportions. The results of the analysis showed that the AI variable was the only one to achieve statistical significance (p < 0.05), indicating its positive impact on warehouse performance. The coefficient of determination for the extended model was R² = 0.061, confirming the validity of its construction. The study indicates that integrating AI with warehouse processes may be a key factor in enhancing operational efficiency and organizational competitiveness. The results obtained justify the need for further research using empirical data and nonlinear models. Model estimation was carried out in the Python environment using the statsmodels library.

11:30-13:30 Session 15D: T5: Digital Transformation 4
Location: Room 4
11:30
Leveraging machine learning techniques for discovering broken lineage links between database objects
PRESENTER: Paweł Boiński

ABSTRACT. Data lineage is the set of techniques for tracking the flow of data throughout its lifecycle. These techniques are crucial for data management, governance, and compliance with regulations. Lineage links are maintained between data and database objects, but they are often broken by temporary objects and user defined functions. To the best of our knowledge, discovering broken lineage links has not been addressed yet in research. In this paper, we present a method for detecting broken lineage links between database objects. To this end we apply machine learning techniques on available metadata. We extract feature vectors and employ a classification approach to determine whether one database object is a source for another. Initial experiments on large database schemas show that the discovery of broken lineage links is possible at an acceptably high probability.

11:50
Integration of IT Systems Using QR Code and IPA in Logistics Process Management

ABSTRACT. In the face of the growing complexity of supply chains and the variety of IT systems used in enterprises, ensuring effective information exchange is a challenge for companies. The research problem addressed in this work is the lack of an effective mechanism for automatic identification and matching data between systems. Therefore, the concept of using QR code technology and IPA systems was proposed. A case study was used to verify the concept. The implementation showed precise product identification, automatic document generation and delivery status update.

12:10
Social challenges and barriers in implementing AI chatbots as part of customer service digital transformation

ABSTRACT. The rapid advancement of artificial intelligence (AI) significantly reshapes customer service, posing notable social challenges and barriers within digital transformation. This study explores user perceptions and societal resistance toward AI-driven chatbots based on a large-scale survey of 11,628 respondents from Poland, Italy, and Sweden. Using a structured framework, chatbot attributes were categorized into essential (response accuracy, real-time assistance), and performance-enhancing (personalization, emotional intelligence, hybrid interactions). Results highlight significant social resistance, especially in emotionally sensitive interactions, where customers strongly prefer human agents over AI. Key barriers identified include privacy concerns, data security risks, and transparency issues. Crucially, trust, explainability, and user education emerged as vital for reducing societal hesitation and fostering acceptance. These findings offer critical insights into social dimensions of digital transformation, emphasizing the importance of developing hybrid customer service models that effectively balance automated technologies and human interaction to enhance consumer trust and overall service experience.

12:30
Do Social Media Virtual Communities Support Purchase Decision Making?
PRESENTER: Donatien Yeto

ABSTRACT. In this article, we investigate whether virtual communities gathered around influencers can support decision-making in the purchasing process. Using data from viewers' opinions under YouTube Influencer's video, the netnography method, and Simon's rational decision-making theory, we investigate: (i) which phases of the purchasing decision-making process can be supported by virtual communities; (ii) what behaviours, structures, and decision-making sequences are visible across the entire virtual community network studied. The results of our study prove that social media content creators as influencers and their virtual communities can affect all phases of purchasing decision-making.

12:50
Digital Transformation in Automotive: Color Design of Cockpit Alerts for Effective and User-Friendly Driver Communication

ABSTRACT. The effectiveness of human-machine interfaces (HMI) in driving automation systems partially depends on how alerts and requests are delivered to the driver without unnecessarily distracting them from the driving task. Depending on the required user reaction time, alerts should be displayed differently - either capture attention immediately for a short duration or remain visible more subtly over time. Our study focuses on the use of color as a means of enhancing alert visibility and attention capture. Based on a user-centered experiment, we developed a ranking of color combinations for both day and night interface modes. Few existing studies address the adaptation of color to interface mode, and when they do, they primarily focus on contrast rather than differentiating alert types by importance. Our approach introduces an additional criterion - the criticality of the alert - that is a key factor in tailoring the visual design of warning signals.

13:10
Classification of enterprises in terms of barriers and risks resulting from the implementation of cloud computing using ELECTRE TRI

ABSTRACT. The development of cloud computing has opened up new opportunities for enterprises in implementing information and communication technologies. Despite the numerous benefits associated with its use, certain barriers and risks also emerge, affecting organizational functioning. To assess their nature and scope, the authors conducted empirical research among companies using services based on the Cloud Computing model, applying the ELECTRE TRI method — a tool belonging to the group of Multi-Criteria Decision Analysis (MCDA) methods. The analysis of the results indicated that the most significant challenges are of an organizational nature. At the same time, most of the surveyed enterprises were classified into categories indicating a low or moderate level of perceived barriers and risks.

11:30-13:30 Session 15F: Industry Session I1
11:30
Banca Intesa Presentation. TBA

ABSTRACT. Banca Intesa Presentation. TBA.

12:15
Automation by Innovation: How KEBA’s Kemro X Platform Builds the Foundation for Digital Ecosystems in Industrial Automation
PRESENTER: Srđan Usorac

ABSTRACT. „KEBA’s motto, “Automation by Innovation”, drives the development of Kemro X, a modular platform that enables scalable, data-driven industrial ecosystems. Building on this backbone, products like Delem and Drag&Bot deliver cutting-edge solutions for sheet metal processing and intuitive robotics, demonstrating how platform innovation translates into real-world value. This presentation will share how KEBA integrates hardware, software, and data engineering to enable digitalization, interoperability, and future-ready automation solutions.

13:30-14:30Lunch
14:30-16:30 Session 16A: T4: Data Science and Machine Learning 7
Location: Room 1
14:30
Stock Turning Points Prediction Using Convolution Neural Networks with Return Rate Heat Map

ABSTRACT. A turning point in the stock market is the moment when the stock price time series changes direction. Its identification is important both for economic theory and, above all, for the practice of investing in the stock market. In this article, we propose a new concept, based on a broad representation of the market in the form of heat maps visualizing histograms of return rates. We then use convolution neural networks (CNN) to process these images. By extracting and analyzing image-based features, we can classify stock market events and effectively detect potential turning points. We conduct the study on 4 datasets, verifying 9 CNN architectures. The results are promising, we achieved accuracy of up to 97%.

14:50
Annotator-Aware Evidential Learning for Polish Clinical Sentences

ABSTRACT. Automatic coding of Polish clinical text is still under-explored. We therefore benchmark six transformers—DISTILBERT, HERBERT, POLBERT, POLKA 1.1B CHAT, XLM-ROBERTA and PAPUGAPT2—on 31 034 de-identified phrases grouped into six clinical categories. A unified fine-tuning and statistically rigorous pipeline (Kruskal–Wallis/ANOVA, Holm-corrected Dunn tests, Cliff’s δ, bootstrap CIs) over 186 210 predictions shows that architecture alone ex- plains∼96% of the variance in top-1 confidence (H = 1.7 ×105, p<10−300). Multilingual XLM-ROBERTA leads; only POLBERT overlaps meaningfully (|δ| = 0.30), whereas all other pairs are near-maximally separated (|δ| > 0.90). The top-quartile confidence peaks at 0.68—be- low clinical automation thresholds—highlighting the need for domain-specific pre-training and macro-F1 evaluation. Open-source code and templates make the benchmark fully reproducible and extensible for Polish biomedical NLP.

15:10
Reconstruction of Images Transmitted over LoRa Networks Using Fully Convolutional Neural Network

ABSTRACT. LoRaWAN networks, which are extremely popular today, are based on the LoRa protocol and offer very long communication ranges, but they also come with significant limitations. These limitations stem primarily from two factors: duty cycle and maximum message size. During image transmission over LoRa-based networks, packet loss is a common problem resulting from limited bandwidth and transmission interference. Another issue is packet reordering, as packets often arrive at the base station out of order compared to their transmission time. It is possible to add metadata (e.g., packet ID or checksum) to the header, but this reduces the space available for actual image data. During image transmission, it leads to missing data in the received content, most often visible as vertical or horizontal lines. In this paper, we present a method to repair such corrupted images using a fully convolutional neural network inspired by the U-Net architecture. The conducted experiments show that the proposed approach effectively reconstructs missing parts of the image, achieving high structural similarity (SSIM) with the original. The proposed method can be successfully applied to image transmission and reconstruction on low-power devices, typical of IoT systems using LoRa for communication.

15:30
End-User Challenges in Implementing Metaheuristic Algorithms from Scientific Papers

ABSTRACT. This preliminary study investigates how scientific paper components influence end-users' metaheuristic algorithms (MAs) implementation. Focusing on two papers presenting the Monkey King Evolution (MKE) and Sine Cosine Algorithm (SCA), we examined a sample of 22 Computer Science engineers tasked with implementing these MAs based on the papers and supplementary resources. Through questionnaires and statistical analysis, we assessed the impact of pseudo-code, Figures, Tables, and overall presentation on comprehension and implementation success. The findings indicate that coherent pseudo-code and visual aids, such as Figures illustrating solution movements, are critical for end-users, while inaccuracies or complex descriptions hinder implementation. The SCA paper, rated significantly higher than the MKE paper, facilitated faster and more accurate implementations. Despite access to the MATLAB source code, challenges persisted in adapting the implementations to Java, highlighting the need for platform-agnostic resources. We propose guidelines for MA papers, including detailed pseudo-code, diverse implementations, and additional implementation instructions. Although limited by a small sample, these insights justify further research into improving scientific papers for practical MA adoption.

14:30-16:30 Session 16B: T7: New Topics in IS Development 2
Location: Room 2
14:30
A fusion of ancient metatecture™ with novelty IS architecture

ABSTRACT. This paper is a preliminary description of Information System for empowering indigenous communities. The system uses a Metaverse interface with digital Marae. The Marae Atea is a sacred space, and the Marae building itself is a kind of knowledge base with community governance rules. The Marae Atea is a pivotal point around which the complete community life is revolving. There is a notifiable resemblance between Marae governance and DAO (Decentralised Autonomous Organization) governance, however Marae concept is thousand years earlier than DAO. The paper describes the information system employed in a sensitive cultural context of indigenous community. Technically, the system is based on blockchain concepts with user interface in Metaverse.

14:50
Let Me Entertain You: On the Bias of Editorial and Algorithmic Recommendations in Public Service Media

ABSTRACT. In today’s online world, bias is ubiquitous. In a Video-on-Demand (VOD) context, there are two main sources for bias: (1) content providers and their methods that generate recommendations; (2) user preferences. We use real world data from one of the largest German Public Service Media (PSM) platforms to study both possible causes by examining content recommendations and content consumption (classified as Education, Information, Culture, and Entertainment according to the regulations governing PSM operations). We find that Education and Entertainment content to be recommended disproportionately. Algorithmic personalized recommendations lead to a bias towards both Education and Entertainment for the recommendations. Editorial content recommendations favor Entertainment content. Additionally, we find that the bias introduced by the recommender systems to be larger than the one introduced by the viewer. We recommend testing an increased use of algorithmic recommendations as these might lead to a lower bias towards Entertainment content.

15:10
Cloud Computing and Its Role in Enterprise Management and Innovation: Insights from Poland

ABSTRACT. Cloud computing (CC) plays a key role in the digital transformation of SMEs by providing flexible IT resources that support management and innovation. This study aimed to assess the impact of CC on innovation and management in over 400 Polish SMEs. The results show that CC particularly supports resource management in medium-sized and commercial enterprises and facilitates innovation in companies with higher profits, greater investments, and larger customer bases. Research limitations, such as the overrepresentation of certain industries, suggest caution in generalizing the findings. Nevertheless, CC emerges as a crucial factor for SME growth and competitiveness, indicating the need for further in-depth research.

15:30
Trust in the Unknown: Bridging Socio-Technical Gaps in the Trustworthiness of Emerging Electronic Identification Systems

ABSTRACT. This paper analyzes stakeholders’ understanding of an electronic identification (e-ID) system based on artificial intelligence and distributed ledger technology. We address the question "How is the trustworthiness of a novel information system for e-ID influenced by the stakeholders’ understanding of its base technologies?". Our findings are based on a qualitative analysis of a questionnaire and interviews with stakeholders from a system development project focused on e-ID for online public services. We found that current e-ID systems have good usability but lack dialog and feedback mechanisms, whereas technical robustness and data protection are deemed essential attributes of emerging solutions. We identified four profiles of prospective users according to variations of trust in the new e-ID system. These findings suggest the need for greater transparency to facilitate the adoption of nascent digital identity solutions.

14:30-16:30 Session 16C: T8: Business Models and IS for Smart Environments 2
Location: Room 3
14:30
Micromobility in a smart environment - a spatial approach

ABSTRACT. This paper is concerned with the role of a new business model based on micromobility, which is part of the concept of intelligent environmental management. An analysis of the subject, the available literature and published research has shown that there is a research gap, with an apparent lack of research into the users of micromobility solutions, particularly in the important aspect of daily school commuting and adult use. The authors present the results of a study on perceptions of micromobility and preferences of adults and primary and secondary school students in this area. The article shows the responsiveness of adults to proposed micromobility solutions. In addition, special attention was given to research on how to get to and from school among primary and secondary school students. The research has helped to reveal the role of and differences in attitudes towards micromobility among urban and rural residents. Furthermore, the authors identified and described the barriers restricting the development of micromobility and provided recommendations for its further development.

14:50
CROWDSOURCING AS A SOCIAL PARTNER OF TRANSDISCIPLINARY PROJECTS

ABSTRACT. The paper explores the usage of crowdsourcing (CS) as a social partner in transdisciplinary (TD) projects, highlighting its value in stakeholder engagement, knowledge integration and social impact in sustainable development and innovation. A two-stage study was conducted, including a literature review and concept-centric analysis of 11 projects, to examine the role of CS throughout the TD project lifecycle. The research introduces a maturity model that systematises CS involvement by identifying different levels of engagement. The findings show that CS enhances the co-production of solutions, improves outcomes and broadens their dissemination through societal participation. The study provides a practical framework (model 4L) for researchers and policymakers to integrate CS into TD projects, thereby promoting social innovation and effective responses to complex challenges.

14:30-16:30 Session 16D: T5: Digital Transformation 5
Location: Room 4
14:30
Perception of FinTech Products and Services Among Rural and Urban Residents

ABSTRACT. This study compares the perceptions of FinTech services among rural and urban residents in Poland. It draws on data from a CAWI survey conducted in 2020 (N=1,153) and uses chi-squared tests, the Fisher-Freeman-Halton test, and post-hoc comparisons. The results indicate that while general attitudes towards FinTech do not differ significantly between rural and urban residents, rural respondents without prior FinTech experience are less likely to consider security and offer attractiveness as key decision factors than residents of small cities. Conversely, device ownership and social influence were less relevant for rural residents than those in mid-sized cities. No significant differences in service evaluation were observed among experienced FinTech users. These findings suggest that perceived differences primarily concern the decision-making stage, not actual use. The results may support the adaptation of FinTech development strategies to the needs of different social groups, thereby enhancing financial inclusion.

14:50
Application of Hybrid Systems SARIMA ANFIS for Monitoring Workforce Dynamics

ABSTRACT. This paper uses a hybrid SARIMA system and an adaptive neuro-fuzzy inference system (ANFIS) to analyze data on interactions between employees in a real-world entity, including email exchanges, chat messages from meetings, and in-person meetings, for the purpose of detecting position changes such as promotions, demotions, and supervisor changes. The dataset, comprising approximately 184 GB of textual data, includes sixteen features related to employee interactions, such as internal contacts, communication with supervisors, subordinates, and individuals at various levels of the hierarchy. The developed system achieved a detection accuracy of 96%, confirming its usefulness in monitoring personnel processes and optimizing human resource management. In this study, the SARIMA model was combined with the ANFIS system, enabling more precise forecasting of changes over time and the detection of employee behaviors, such as sudden position changes or team interactions. By operating in a quasi-real-time mode, the system allows for the rapid identification of potential irregularities, enhancing organizational security and supporting personnel decision-making in dynamically changing conditions. The results of our research indicate that hybrid models integrating the analysis of large datasets and flexible inference systems can effectively support management and behavioral profiling in organizations.

15:10
Dynamic generation of Decision Model and Notation rules for tax regulations – case study from Swiss accounting offices
PRESENTER: Marcin Makowski

ABSTRACT. This study investigates the integration of Decision Model and Notation (DMN) with generative artificial intelligence (AI) to support the dynamic and explainable automation of decision-making processes in the domain of tax-related accounting. Addressing a recognized gap in the literature, we propose and empirically validate a hybrid human-AI framework for decision rule generation, implemented within LUCA - an enterprise-grade system deployed across Swiss accounting offices. The system leverages BPMN-driven workflows, AI-supported document understanding, and DMN-based rule management to formalize tacit expert knowledge. The methodology is grounded in a descriptive case study approach, utilizing participant observation and iterative validation. Key findings indicate a substantial improvement in automation efficiency, enhanced auditability, and effective rule reuse across distributed organizational units. We demonstrate that combining generative AI with DMN enables scalable, high-fidelity decision modeling while preserving transparency and human oversight. The study contributes a replicable model of knowledge acquisition and governance for data- and regulation-intensive environments.

15:30
The Interplay of R&D, Digitalisation, and GDP: Insights from Fuzzy Logic Analysis

ABSTRACT. Technological development is recognized as a key factor in long-term economic growth, and R&D activity is a key driver of this process. In the context of the ongoing digital transformation, it is necessary to understand the links between R&D spending, the level of digitization and the level of GDP per capita in an integrated manner. While existing studies examine these variables separately using classical statistical methods, they fail to capture the indirect and uncertain relationships between them. Therefore, this paper aims to develop a model of the relationship between R&D expenditures, DESI level and GDP per capita using fuzzy logic. The practical application of the method is presented on the example of the EU-27 countries. The results prove the usefulness of the method for analyzing real causal relationships. The model based on fuzzy logic made it possible to identify differences between countries that could go unnoticed using traditional classification methods.

15:50
AI as an element of digital transformation in the activities of critical infrastructure regulators in the context of the personality and behavioural conditions of officials

ABSTRACT. The aim of this article is to identify personality and behavioural conditions as determinants of AI implementation by critical infrastructure (telecommunication and energy) regulators in Poland. To achieve the research objective the results of a survey and experimental research were used. They were conducted between 2022 and 2023 among officials of the Office of Electronic Communications and the Energy Regulatory Office. It was shown that officials are responsible optimists willing to cooperate. They are individuals with relatively low levels of trust in others as well as moderate assertiveness and low risk aversion. Their risk propensity is consistent with prospect theory. They perceive more risk in making decisions under uncertainty than under risk. In contrast officials' succumbing to the status quo effect and the sunk cost effect is dependent on framing effect. Personality and behavioural conditions have implications for the implementation of AI by regulatory authorities.

16:10
Key Factors Determining Decisions on the Implementation of Robotic Process Automation

ABSTRACT. The article presents the results of research into the identification of key factors influencing decisions on the implementation of Robotic Process Automation (RPA) in enterprises. The study uses the theoretical framework of Davis' Technology Acceptance Model (TAM) and its modifications which apply to the study of acceptance of new technologies. The study results indicate a direct impact of the following variables: Perceived Ease of Use and Perceived Usefulness on Behavioural Intention. They also confirm that the Perceived Usefulness variable is statistically significantly influenced by the Social Influence, Price Value, and Process Adequacy variables. For the Perceived Ease of Use variable, a statistically significant effect of the User Involvement and Facilitating Conditions variables was presented, and no statistically significant effect of the Implementation Method variable was demonstrated. The obtained results contribute to reducing the indicated research gap concerning the identification of factors that are very important for the implementation of RPA.

14:30-16:30 Session 16F: Industry Session I2
14:30
EGZAKTA Group Presentation. TBA

ABSTRACT. EGZAKTA Group Presentation. TBA.

15:15
ASEE Presentation. TBA

ABSTRACT. ASEE Presentation. TBA.

17:30-23:30Excursion + Dinner (Outside Belgrade)

Buses will start in front of the faculty building, on the side of Bulevar oslobođenja street.