ABSTRACT. Disruptive ecosystems lead to new dimensions also in IS development and operation, which poses a significant challenge in managing change across multi-disciplinary teams and leading innovation initiatives at a global level. Design thinking is introduced to tackle this complex task, as it applies problem-solving techniques through co-creation among stakeholders. It enables early exploration and validation of design(s) of new services and disruptive business models. The Scene2Model tool allows the creation of the digital twin of the design thinking artifacts to be shared among globally distributed stakeholders and IS developers, and maximize collective intelligence efforts in the digital age. In this context, the digital twins are digital conceptual models that can be further enriched with domain knowledge to be integrated with existing business assets. The interplay of conceptual modelling and design thinking establishes a connection between unrestrained design artifacts and more formal abstractions, facilitating digital agility and innovation in IS development.

ABSTRACT. The aim of the research was to determine the comprehensibility of the content of these local authority websites and the level of public perception of them. Methods of sampling, analysis and comparison were used. It was found that citizens with a low level of education or who have reading difficulties have a limited ability to understand and perceive the material on government websites. This deprives them of the opportunity to use electronic government services. The study revealed some of the factors influencing citizens' perception of government online content and showed that the readability of information on websites plays an important role. As a result, it was found that widening the circle of Poles as users of electronic services is possible by editing content with special software. This is necessary in order to simplify them and thus increase the level of perception of online materials by different categories of citizens.

ABSTRACT. Smart Cities are urban environments that use digital technology and data-driven solutions to improve citizens' efficiency, sustainability, and quality of life, especially using Artificial Intelligence to improve human lives. These methods can also help urban residents in dangerous situations by detecting dangerous situations and securing communication during emergency services notifications. In this paper, we present the security protocol that ensures secure communication during the notification of emergency services. This protocol is secure, lightweight, and scalable. We verified its security using an automated tool that did not detect a protocol attack.

ABSTRACT. Ensuring optimal functioning and efficiency of offshore wind farms requires effective performance maintenance strategies. This paper introduces a novel, innovative treatment of the problem in the context of a comprehensive strategy to optimize preventive maintenance for wind farms. The developed declarative reference model allows capturing the trade-off between the costs of losses incurred due to extending the servicing time of wind turbines and the expenditure incurred on the scale of the fleet of vessels used and the service teams employed to maintain them. The problem of meeting constraints formulated in this representation allows for identifying and mitigating potential problems before they occur. It is implemented in a declarative programming environment (IBM ILOG CPLEX Optimization Studio), creating a prototype of a task-oriented Decision Support System, confirmed by several experiments.

ABSTRACT. Various metrics exist for evaluating UML diagrams, including entropy-based ones like ours, which assess information content. They allow for judging certain features of the design that depend on the information content. This paper proposes the FBS24 use case diagram measure, which should ensure that the software architecture design consists of mature UML diagrams. Detecting an inappropriate (immature, unfinished) Use Case Diagram before the software development phase can stall the entire software development process until a mature UCD is developed. Currently, no indicators (metrics) show the maturity (or lack of applicability) of diagrams. Moreover, in most such software metrics, weights are selected arbitrarily, which leads to numerous anomalies. We show how to construct the measure most resistant to such anomalies. We also show how to check the correctness and usefulness of the constructed measure. As our measure is objective, it is suitable for remote work of distributed teams building IT systems.

ABSTRACT. Convolutional Neural Networks (CNNs) are widely used in computer vision, which is now increasingly used in mobile phones. The problem is that smartphones do not have much processing power. Initially, CNNs focused solely on increasing accuracy. High-end computing devices are most often used in this type of research. The most popular application of lightweight CNN object detection is real-time image processing, which can be found in devices such as cameras and autonomous vehicles. Therefore, there is a need to optimize CNNs for use on mobile devices. This paper presents the comparision of latency and mAP of 22 lightweight CNN models from the MobileNet and EfficientDet families measured on 7 mobile phones.

ABSTRACT. Research shows results and describes solution and techniques allowing for conversion of highly detailed real world maps into simpler maps which can be used in real-time 3D mobile games focusing mainly on medium and low computing power devices. This approach is especially interesting for mobile 3D apps developers working with real world maps that can be implemented in games like tycoons, strategies or geolocation type. The obtained results show possibility of achieving good compromise between level of details extracted from original maps and high performance of 3D graphics generated on various mobile devices.

ABSTRACT. The article focuses on the topic of remote work, which has permanently become part of the economic landscape with the emergence of the threat posed by Covid-19. The article presents the assumptions and research methodology of the ongoing scientific project aimed at developing an effective remote work management model. Due to the interdisciplinary and turbulent organizational-functional environment of remote work, and in order to maintain high quality and universality of results, mixed-method research has been decided upon. The research procedure has been meticulously described and presented graphically. Additionally, tools that have been or will be used in the process of statistical inference have been indicated. To enhance understanding of the complexity of the research problem, the article includes a theoretical construct that serves as the foundation for the entire project. This construct encompasses forces and areas identified through thorough literature review, which influence and condition effective management of remote work.

ABSTRACT. The complexity of data in smart cities creates challenges for developers and hinders cohesive understanding of diverse datasets. These critical data sets are often underutilized due to opaque organization and accessibility. In our research, we use Large Language Models (LLMs) as "data custodians" to improve data navigability and usability in smart city platforms. We evaluated the ability of LLMs to generate accurate data descriptions, identify feature names, and discern relationships from limited raw data, demonstrating their proficiency with minimal input

ABSTRACT. This paper presents an introductory study of professional music software after-sales service, in particular, user feedback and software evolution perception, to evolve it as part of the system development process. It has been based on theoretical and managerial review. During the first stage, a new conceptual model of user feedback based on the agile software development cycle integrating elements of TAM, CX, UX, and Agile PM methodologies was prepared. At this entry stage, qualitative research was chosen. Data were collected through semi-structured hybrid interviews among professional music software users. The outcomes may be a valuable source of ideas on professional users’ software development process improvement in the music industry and on the pace of introducing changes and updates to the software. Results may be valuable for both developers and researchers. Overall, practical changes based on our introductory findings may improve customer satisfaction in this market.

ABSTRACT. The aim of the article is to learn about travelers' behavior in the use of modern technologies in transport in the context of changes taking place in the environment. The research adopted a hypothesis: the "digital" development of passenger behavior coincided with the unexpectedly rapid development of modern technologies and their "forced" use during the COVID-19 pandemic. The article presents the results of four original studies conducted in 2020-2023 using the CAWI method on representative nationwide samples: N = 674 (2020), 890 (2021), 364 (2022), 929 (2023). Respondents were asked whether they use modern technologies to obtain information, purchase tickets and check the route or timetable in passenger transport. In 2020, 53% answered affirmatively and 47% declared not to use modern technologies. In 2023, the group of users increased to 78% and the group of non-users decreased to 22%.

ABSTRACT. The paper describes immersive capabilities of CAVE Automatic Virtual Environment and Head-Mounted Display. An important aspect of this research was to develop a method for quantifying user immersion in both systems. Two virtual reality applications, "Flat of Negative Emotions" and "Arachnophobia Treatment Support" were used to observe and analyze user reactions and engagement levels. Participants were exposed to these applica¬tions in both environments, allowing for a comparative analysis of the technologies. The methodology incorporated surveys, observation forms, and direct interaction analysis, combining qualitative and quantitative data for a comprehensive evaluation of both sys¬tems. The study involved 124 participants from varied backgrounds. The paper presents the objectives, methodology, and findings, with a focus on comparing the immersion levels in CAVE and HMD settings. The results contribute to the academic discourse in virtual reality and human-computer interaction, offering methodological advancements in measuring immersion and guiding future research in immersive technology.

ABSTRACT. The use of many different types of data sensors makes it possible to better represent and understand a given phenomenon. However, the problem becomes the synchronization and fusion of this data. Our goal was to develop a lightweight and flexible system for synchronized data acquisition from various sensors. We designed the Synchronized Data Acquisition System (SDAS), which uses a self-designed Edge Control Protocol (ECP) and Temporal Sample Alignment (TSA) algorithm to synchronize the acquired samples across all the sensors connected to the SDAS. As samples are synchronized during writing data files, we can call that a software-based synchronization. We also conducted tests to validate our solution.

ABSTRACT. A few decades ago, insurance policies were manually processed, with prices calculated using calculators, and customer service conducted at home. Currently, insurers utilize modern IT systems, and the insurance industry is thriving. The latest technological advancements, such as telematics (for assessing driving risk), artificial intelligence (for automatic insurance claim evaluation), and machine learning (for creating insurance products safeguarding businesses from weather anomalies), are employed. Generation Z individuals are entering the job market, including the insurance sector. Their needs regarding various system uses, extensive tech experience, self-awareness, and responsible data protection, must be considered. This article offers a brief overview and potential future directions for IT systems in insurance, tailored to Generation Z's needs

ABSTRACT. On the one hand, in recent years, end-to-end solutions for running various data engineering experiments and getting insights from them are gaining interest from research communities. The insights are often learned from applying machine learning algorithms on experimental data. In this context, experiments repeatability and open access experimental data become new important trends. On the other hand, with the widespread of big data, integration architectures and processes are among the most frequently researched topics, as they are critical in modern data management systems, aimed at consolidating data from diverse sources to offer a unified perspective for users. In this paper we propose an end-to-end prototype system for collecting and analyzing performance characteristics of code snippets. The system was built, deployed, and tested for the problem of building performance characteristics of user defined functions.

ABSTRACT. The article presents the results of a quantitative study of a group of micro-enterprises in Poland's industrial sector in the digital transformation field, considering the COVID-19 pandemic. The authors aimed to check the status quo of implementing digital solutions and technological innovations in business. Concluding, Polish micro-enterprises need more time to use digital solutions, implement technological innovations, and achieve digital transformation. The COVID-19 pandemic period, to some extent, gave rise to the need to implement digital solutions, but it did not happen suddenly and on a large scale. Advanced technological innovations such as virtual reality, augmented reality, robotization, or automation of production processes are still not fully understood by the entities surveyed and are even less of interest to them. Companies do not look for sources of financing and support from external entities to implement digital solutions for the company.

ABSTRACT. This study introduces a novel method to boost grant application success by analyzing the convergence between project abstracts and funding call texts using text mining. Focusing on Horizon 2020 data, it aims to identify how textual alignment affects funding outcomes and establish a quantifiable similarity threshold for eligibility. The research proposes a strategic framework to improve grant proposal competitiveness, especially benefiting institutions in the European Union with lower success rates, by providing a practical tool for enhancing grant writing processes. The findings are expected to fill a literature gap, offering empirical evidence on the relationship between textual similarity and funding success. This research not only has the potential to guide future grant writing practices but also provides guidelines for the internal (within the organization) application preselection, aiming to increase the effectiveness in obtaining research funds.

ABSTRACT. This paper evaluates the \texttt{gpt-4-turbo} model’s proficiency in recognizing named entities within the clinical trial eligibility criteria. We employ prompt learning to a dataset comprising $49\,903$ criteria from $3\,314$ trials, with $120\,906$ annotated entities in 15 classes. We compare the performance of \texttt{gpt-4-turbo} to state-of-the-art BERT-based Transformer models\footnote{Due to page limits, detailed results and code listings are presented in the supplementary material available at https://github.com/megaduks/isd24}. Contrary to expectations, BERT-based models outperform \texttt{gpt-4-turbo} after moderate fine-tuning, in particular in low-resource settings. The \texttt{CODER} model consistently surpasses others in both low- and high-resource environments, likely due to term normalization and extensive pre-training on the UMLS thesaurus. However, it is important to recognize that traditional NER evaluation metrics, such as precision, recall, and the $F_1$ score, can unfairly penalize generative language models, even if they correctly identify entities.

ABSTRACT. Smart contracts are pivotal in blockchain systems, yet ensuring their reliability and security remains challenging due to coding complexities and potential vulnerabilities. This paper explores the use of Large Language Models (LLMs) in enhancing the smart contract code quality. As part of leveraging extensive training data and language understanding, we experiment with different approaches. LLMs aid developers by offering automated code suggestions, identifying vulnerabilities and promoting best practices. Through experimentation, we demonstrate how integrating LLM-based approaches improves code quality and reliability in blockchain applications.

ABSTRACT. This article contains a design scheme and functional characteristics of a decision support system aimed at analysing high-frequency ball tracking data and deriving real-time recommendations to players of individual games like golf or mini golf. The characteristic feature of such game is the lack of other players’ responses to a player's move. The game becomes competitive when one of players completes certain task first or the final scores of players are compared. We review the best practices of game support methods and propose the software architecture of a mini-golf game support system (GSS) to be implemented by a course manager or supplier. The system is based on real-time analysis of high-frequency data captured during the ball movement. It uses a dynamic programming-type algorithm to compute optimal moves from the current position of the player and provides hints accounting for the hitherto score and game level of this player.

ABSTRACT. The emergence and development of Cloud Computing (CC) provide enterprises with new opportunities in acquiring and using information and communication technologies. Implementing cloud computing can benefit enterprises in several aspects, i.e., strategic, economic, organizational, technological, social, and environmental. To assess the diversity of benefits resulting from using CC in an enterprise, the authors conducted empirical research among enterprises using services available in the CC in their operations. The results were assessed using the ELECTRE TRI, one of the methods from the ELECTRE family included in the Multi-Criteria Decision Analysis group. The final part of the article presents the results of the conducted analysis. The results indicated that the benefits from the organizational aspect were rated the highest. However, in general terms (all benefits), most enterprises belong to classes 1-2 and 2, which indicates that the examined enterprise evaluates the benefits resulting from using CC as low or average.

ABSTRACT. Supply Chain 4.0 is the result of the fourth industrial revolution (Industry 4.0). The use of modern technologies and digitalization have become a fact in logistics activities. The development of the Supply Chain 4.0 concept poses a number of new challenges, especially for manufacturing companies that must ensure the efficiency and effectiveness of their processes, also in the area of supply chain management (SCM). The article presents the results of a study aimed at reflecting on the possibilities of investing in modern ICT solutions and digital Supply Chain 4.0 supporting supply chain management processes in Polish and German micro, small and medium-sized enterprises.

ABSTRACT. Modern cities, against global plans promoting sustainability, are still being designed and built with a primary focus on the needs of drivers. Planning concepts, such as a 15-minute city, aim to minimise car usage by assuring quick access to vital urban functions within walking distance. However, their application needs information about achievability and viability. This work presents a model that combines qualitative and quantitative studies on travel duration from home to school and work. The survey data are modelling to present natural citizens' behaviour in their transportation environment. The actual locations respondents visited are used to model their travels and calculate travel parameters. The study performed among parents from three primary schools showed that over 56% of travel to schools can be covered by public transport in less than 15 minutes and that the benefits of using a car on longer travel to work are limited.

ABSTRACT. This study investigates the role of deep learning models, particularly MobileNet-v2, in Parkinson's Disease (PD) detection through handwriting spiral analysis. Handwriting difficulties often signal early signs of PD, necessitating early detection tools due to potential impacts on patients' work capacities. The study utilizes a three-fold approach, including data augmentation, algorithm development for simulated PD image datasets, and the creation of a hybrid dataset. MobileNet-v2 is trained on these datasets, revealing higher generalization or prediction accuracy of 84\% with hybrid datasets. Future research will explore the impact of high variability synthetic datasets on prediction accuracies and investigate the MobileNet-v2 architecture's memory footprint for timely inferences with low latency.

ABSTRACT. In this paper, we address the problem of preparing data for training detectors to identify transparent objects in light microscopy images. To this end, we propose using blends of reference images and monitoring background, instead of time-consuming labelling of monitoring data. This approach allowed us to avoid the need to involve a palynologist in the preparation of the training data while also ensuring 100% correct ground-truth labels. The statistical analysis of the deep learning results confirms that the results obtained for blends only are more stable, and in some cases surpass the results obtained for the training set with some labelled monitoring data added to reference images and monitoring background.

ABSTRACT. Unsupervised models are becoming increasingly common in business processes. They are extremely effective in cases where we don't have a clearly defined decision class or the data contains anomalies that are hard to identify. The problem emerges in the effective processing of categorical data. Recently, many new approaches have been designed to analyze this data type. Still, most of them do not address the issue of unbalanced datasets, which is extremely difficult to catch when dealing with unlabeled data. Moreover, it is sometimes challenging to determine when abnormal observations represent a small data cluster and when they are already anomalies. This research analyzes several less popular algorithms that solve this problem and automatically place abnormal observations into separated clusters. We have shown that such methods are much better at clustering unbalanced data but also perfectly detect outliers in categorical datasets.

ABSTRACT. This study investigates the dynamics of Enterprise Resource Planning (ERP) implementation within contemporary enterprises. Through a comprehensive analysis of organizational perspectives, the research explores the decision-making processes, key considerations, and challenges encountered during the first phase of the ERP adoption. Drawing upon insights from industry professionals and stakeholders, the study delineates strategic approaches to ERP implementation, highlighting the interplay between organizational maturity, system functionality, and integration capabilities. Findings underscore the importance of aligning ERP implementation with established business models while accommodating future scalability and evolving operational needs and the reliability of the vendor which significantly reduces the risk of implementation failure.

ABSTRACT. While prior studies extensively explored digital transformation in for-profit and public organisations, the challenges faced by nonprofits remain under-investigated. This study attempts to bridge this gap in the literature by conducting an exploratory case study within a nonprofit organisation in Sweden. Semi-structured interviews and internal document reviews were used to collect data. The thematic analysis revealed fifteen distinct barriers that hinder nonprofit organisations from realising the anticipated results of digital transformation. The findings provide a foundation for further research on digital transformation in organisations engaged in the nonprofit industry. Moreover, practitioners can also leverage these insights to inform their own digital transformation journeys.

ABSTRACT. In the contemporary, knowledge-based economy, enterprises are forced to bear the costs related to cybersecurity. While breaches negatively affect companies' budgets, accurate decisions on security investments result in visible savings. At the same time, cybersecurity cost assessment methods that support these decisions are lacking. Caspea addresses the gap by enabling the estimation of costs related to personnel activities involved in cybersecurity management. In this paper, new advancements in the research related to the construction of an ISO/IEC 27001-based costing model are described. This includes revising cost centres based on the ISO27k RASCI matrix, minimising input and output data, or implementing a new calculation spreadsheet that contains substantial changes compared to its previous editions. A comparative analysis with the earlier version of Caspea has been performed. The application of the new model to a woodworking company is illustrated. The results show gradual extension and the broader scope of the Caspea framework.

ABSTRACT. Cloud servers are becoming more widely used due to the growing number of mobile devices with limited capabilities for complex computing, processing, and storage. Choosing the optimal cloud server is challenging due to the continuous technological development of the growing number of cloud service providers. The need to evaluate cloud services according to multiple attributes suggests that multi-criteria decision analysis (MCDA) methods are appropriate. This paper proposes an approach to multi-criteria assessment of cloud services using the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) method employing various distance metrics with different approaches to prefer sustainable solutions and criteria compensation. A sensitivity analysis that considered changing criteria weights was applied to assess the robustness of evaluated solutions. The demonstrated approach proved its applicability for multi-criteria cloud service assessment considering sustainability and robustness.

ABSTRACT. The current paper investigates the role of a stakeholder group in the SaaS adoption decision. In the research approach adopted, two stakeholder groups have been examined: SaaS providers and SaaS adopters. Drawing from the opinions of 134 providers and 137 adopters and employing a multifaceted model explaining antecedents of the propensity to adopt SaaS, the current study examines the moderating role of a stakeholder group for the relationships between constructs. The adopted research approach is Partial Least Squares Multigroup Analysis (PLS-MGA). The preliminary results obtained in the PLS-MGA analysis demonstrate that SaaS providers and adopters differ significantly in their decision on SaaS adoption with respect to IT-business alignment and productivity. In particular, adopters appear to underestimate the role of productivity, while providers tend to underestimate the role of IT-business alignment.

ABSTRACT. The paper proposes CQFAU, a new questionnaire for usability assessment, designed for simplicity yet capable of obtaining information on improvement directions suggested by respondents. It differs from the available questionnaires having similar purpose in its scope (focusing at specific function of the software rather than its general impression), question form (using a set of binary questions followed by a single open one), and the cascading structure (presenting further questions only to respondents who are considered knowledgeable to answer them). The paper describes the instrument, explains how it can be used to identify flaws in software usability and for benchmarking usability, and reports the results of its preliminary evaluation.

ABSTRACT. Artificial intelligence and natural language processing are rapidly developing fields. Natural Language Processing (NLP), like other Machine Learning (ML), Deep Learning (DL), and data processing tasks, requires a large amount of data to be effective. Despite the emergence of newer and better NLP models, text processing in languages other than English, such as Polish, is still problematic. Applications that recognize emotion or sarcasm in texts, among others, as those that could help people with intellectual disabilities face many challenges, from a lack of data in a specific language to a shortage of solutions dedicated to such problems. Responding to this need could be the hybrid model created as part of this research. The paper presents the proposal of hybrid method used for emotion and sarcasm classification in Polish that was based on English dedicated methods together with its implementation and evaluation based on performed experiments using the proposed datasets.

ABSTRACT. This article presents a case study on the development of a biometric voice verification system for an intercom solution, utilizing the DeepSpeaker neural network architecture. Despite the variety of solutions available in the literature, there is a noted lack of evaluations for "text-independent" systems under real conditions and with varying distances between the speaker and the microphone. This article aims to bridge this gap. The study explores the impact of different types of parameterizations on network performance, the effects of signal augmentation, and the results obtained under conditions of low Signal-to-Noise Ratio (SNR) and reverberation. The findings indicate a significant need for further research, as they suggest substantial room for improvement.

ABSTRACT. There are around 50-100 thousand deaf people in Poland, their main language is Polish sign language. It can be challenging for them to communicate with the rest of society and there is a gap in Polish sign language gestures to text conversion. Although some research has been done before, no research paper or product solves this problem. The primary objective is to develop a concept of an intelligent application that can convert Polish sign language from either a video or a live feed. To achieve this, research was conducted on other sign languages, which helped in selecting the most promising hybrid models of deep neural networks. Subsequently, tests were conducted and the best model was chosen. Finally, the best model was trained on the dataset of Polish sign language, using weights (transfer learning) trained on the MS American Sign Language dataset.

ABSTRACT. This paper presents a solution for generating corpora of simulated Polish speech recordings in complex acoustic environments. The proposed method introduces an additional layer of unpredictable sound events, in addition to the acoustic scene noise and reverberation, making the solution unique. We generated a corpus comprising over 277 hours of training examples and over 5.5 hours for testing purposes using publicly available data sources. Next, we trained the Conv-TasNet network on the generated data to enhance single speech and separate two speakers from complex noise. The results of the experiments indicated the potential of the generated corpora for solving these tasks. Researchers can use publicly available codes to create their corpora tailored to the Polish language and solve various speech-related tasks.

ABSTRACT. Given that churn management is a crucial endeavour for firms aiming to retain valuable customers, the capacity to forecast customer churn is indispensable. We use rough and fuzzy set based classifier to predict customer churn on the example of the Bank Customer Churn dataset. Rough set theory offers techniques for handling incomplete or missing data. By utilizing lower and upper approximation concepts, the system can still perform prediction even when certain feature values are missing what we show in the paper for every combination of missing features. Moreover, we determine feature importance coefficient evaluated through two different means: directly from data and from the working classifier. Rough set-based systems can be integrated with other machine learning and data mining techniques, and we use the LEM-2 rule induction algorithm to create a rule base for the rough-fuzzy classifier.

ABSTRACT. Data source connectors are core components of any data integration architecture. Typically, they are deployed as libraries of connectors. Such deployment exposes some significant pitfalls, e.g., poor maintainability, limited scalability, limited performance, and challenging security. To mitigate these drawbacks, we propose to organize connectors as the so-called a library of connectors used as a service (LCS). In this paper, we discuss design patterns of the LCS that allow to ease connectors maintenance, enhance data access security, and increase performance.

ABSTRACT. The Node.js ecosystem features a large number of backend application frameworks with diverse performance-related characteristics. Since performance is a critical factor for modern-world applications, they need to be carefully examined. This paper analyses performance of various frameworks and investigates changes in performance characteristics under sustained load, grounded in the analysis of runtime environment internals. Results of a performance experiment indicate significant differences in application throughput obtained by different Node.js frameworks with a particular characteristic shared between all of them, yet absent in baseline SUTs implemented in other programming languages.

ABSTRACT. This study explores how healthcare providers in South African public hospitals enact workaround practices to an m-health intervention to overcome its limitations and constraints. Employing the work systems method, we analyze the causing factors driving the enactment of workaround practices to an m-health intervention in the Western Cape referral system, in South Africa. A total of 15 semi-structured interviews were conducted with medical officers and IT personnel, to explore the rationales behind their enactment of workarounds. Despite the reported benefits such as improved mobility and communication between medical officers and specialists, the m-health intervention is currently not used as intended. Instead, health care providers (HCPs) are enacting workarounds to the current structures and subverting the m-health intervention due to design-reality gaps. While these design-reality gaps exist, these workarounds also offer opportunities for innovation and process improvements to the patient referral process in the public hospitals.

ABSTRACT. The current elderly care models are challenged by an ageing population and require digital transformation involving many stakeholders, among which technology providers appear under researched. To bridge this research gap, the current study examined the perceptions of technology providers regarding barriers and enablers in two contrasting socioeconomic contexts: Poland and Sweden. The analysis employed the five-dimensional SIM model as an analytical framework and allowed us to achieve a more nuanced understanding of the technology providers’ viewpoint on determinants of technology adoption for the digital transformation of elderly care. Our preliminary findings suggest that technology providers acknowledge their role in the transformation process and are willing to deliver quality solutions; however, they also perceive a number of environmental barriers that need be addressed at the governmental level. In addition, the results imply that socioeconomic context play a role in establishing a supportive environment for technology providers.

ABSTRACT. This paper presents the findings of a research study focused on accounting firms and their adoption of accounting process automation, including Robotic Process Automation (RPA), Optical Character Recognition (OCR), Electronic Data Interchange (EDI), and electronic invoicing systems. The primary goal of the research was to provide an updated insight into the current practices of accounting firms regarding the automation of accounting processes. Structured interview surveys were conducted at the end of 2023, using the Computer-Assisted Web Interview (CAWI) method, involving 24 accounting firms in Poland. The key findings reveal that approximately 70% of documents were received electronically in formats such as pdf, scan, image, or EDI, and nearly 80% of all paper documents provided by clients required subsequent scanning via OCR. This paper contributes to a better comprehension of process automation in accounting firms and offers an insight into the current state of this phenomenon in Poland.

ABSTRACT. The paper addresses the development of Business Intelligence management dashboards for sustainable, smart and resilient city. It highlights benefits and requirements of the use of data visualisations to communicate information to support decision-makers and the local community in understanding of the complex dependencies between city’s different functional areas. The specific aim of the paper is to show the original City’s Fundamental Power Index constructed to evaluate the city from the perspective of implementing the concepts of a sustainable, smart and resilient city. The dashboard based on the index uses the set of data at city level provided by the national authorities of Poland. The comparative analysis required implementation of taxonomic measure of development. The applied method and the results obtained can be used for different cities also in subsequent years to examine the results of strategic management in the implementation of sustainable, smart and resilient development policy.

ABSTRACT. The research involved creating synthetic samples to enrich the training set and improve classification performance. Data generation was a key element of the biometrics gait system based on wearable sensors. The aim of the study was to investigate which parameters of the Long short-term memory–Mixture Density Networks (LSTM–MDN) models would provide the greatest increase in recognition metrics. Validation was conducted for normalized and non-normalized data for a large 100-person dataset. In the first case, the use of synthetic data from VAE-type generative models increased the F1-score from 0.754 to 0.776, while for proposed architectures increased metrics to 0.789. For normalized data, VAE-based models worsened recognition performance. Whereas the proposed model increased the F1-score from a baseline of 0.928 to 0.966. The conducted experiments indicate that generating synthetic data based on MDN models is more profitable in the cases of distribution shift between training and testing set.

ABSTRACT. Presentation Attack Detection (PAD) is crucial in biometric finger vein recognition. The susceptibility of these systems to forged finger vein images is a significant challenge. Existing approaches to mitigate presentation attacks have computational complexity limitations and limited data availability. This study proposed a novel method for identifying presentation attacks in finger vein biometric systems. We have used optimal Gray-Level Co-occurrence Matrix (GLCM) features with the Light-Gradient Boosting Machine (LGBM) classification model. We use statistical texture attributes namely, energy, correlation, and contrast to extract optimal features from counterfeit and authentic finger-vein images. The study investigates cluster-pixel connectivity in finger vein images. Our approach is tested using K-fold cross-validation and compared to existing methods. Results demonstrate that Light-GBM outperforms other classifiers. The proposed classifier achieved low APCER values of 2.73% and 8.80% compared to other classifiers. The use of Light-GBM in addressing presentation attacks in finger vein biometric systems is highly significant.

ABSTRACT. We present a system for detecting fraud related to illegal transmission of telecommunications traffic of voice calls. This phenomenon, called SIM box, can be identified and limited by using Data Mining customer classification models. The results of these models can then be decomposed by Independent Component Analysis into latent source data from which destructive components can be identified. By identifying these components using Beta Divergence, eliminating them and performing the inverse transformation to Independent Component Analysis, we can improve prediction results. The process is organized in several layers, creating a unified Deep Learning System. We demonstrate the effectiveness of the approach in a practical experiment

ABSTRACT. This study presents the development and evaluation of a novel, multi-model AI system designed to train and deliver multiple deep learning models for the classification of focal lesions in thyroid, using ultrasound images. Leveraging a dataset of 484 images, we trained a diverse array of 1300 models encompassing advanced convolutional neural networks, including ResNet, DenseNet and VGG architectures. To minimize random errors, the training dataset was randomly sampled 20 times. The primary objective was to enhance the diagnostic accuracy in distinguishing benign from malignant thyroid nodules through automated analysis. The performance of our models was rigorously assessed, demonstrating promising results with an average area under the curve of 0.86 and sensitivity of 0.85. These findings highlight the significant potential of integrating deep learning techniques with ultrasound imaging to improve the classification of thyroid nodules.