We investigate the efficacy of Kagi chart representations as structured inputs for deep learning models in algorithmic trading. By transforming noisy financial time series into trend-dependent Kagi segments, we capture significant market movements while suppressing stochastic fluctuations. We compare fixed percentage-based and ATR-based reversal thresholds, assessing their impact on data density, training stability, and predictive performance. Our empirical results show that ATR-based thresholds provide more robust performance across daily and 30-min data. The validation on QuantConnect confirms the practical applicability of the approach. Our main contribution is the demonstration that Kagi segment duplets offer a compact, information-dense encoding suitable for neural network models.
The purpose of the article is to define the essence of digital inclusive technologies in the context of implementing universal design in primary education, to study their potential and practical application for teaching literacy to primary school pupils. The study uses the following methods: analysis of scientific and pedagogical literature, synthesis, generalisation and systematisation of theoretical material and practical experience in the use of digital inclusive resources in the educational process to develop reading and writing skills in primary school children. The use of digital inclusive technologies in primary school is based on the principles of accessibility, flexibility and multi-channel communication. These tools provide personalised learning paths, reduce psycho-emotional stress. The use of platforms such as Edugames and Epic!, and the creation of personalised HTML games using artificial intelligence tools (ChatGPT, Google Gemini), makes it possible to differentiate tasks according to the individual needs of each child. These technologies can support student autonomy through synchronous feedback, gamification elements, and tolerance for mistakes, creating a situation of success for every learner in today’s educational environment.
This article examines the role of the STEM concept and programming within the context of secondary vocational education for students in economics-based programs. The study aims to compare the effectiveness of using physical hardware (the micro:bit microcontroller) versus a virtual development environment (the Microsoft MakeCode simulator). The research employs a mixed-methods design, combining a quantitative questionnaire survey of 56 respondents with a qualitative analysis of semi-structured interviews. The results reveal the existence of a so-called “haptic paradox.” Physical hardware demonstrates significantly higher motivational potential and strengthens the perceived meaningfulness of the work by providing students with tangible evidence of their digital competence. Conversely, the virtual simulator proved to be a more didactically effective tool for developing procedural knowledge and understanding abstract algorithmic structures, primarily due to the elimination of technical barriers and a lower cognitive load during debugging. Based on the findings, the study proposes the implementation of a hybrid instructional model. This model consists of the initial use of a simulator as a cognitive “scaffolding” tool, followed by work with physical devices to anchor attitudes toward technology and bridge the gap between code and real-world practice. This paper provides insight into programming instructional methodologies for a specific target group of students for whom computer science is not a core major subject.
The rapid integration of digital technologies into primary education has intensified the need to examine not only technological tools themselves, but also how future teachers reason about their didactic use. In the field of technical education, 3D pens represent an accessible bridge between creative construction, embodied learning, and engineering-oriented STEAM approaches. However, limited research has focused on how pre-service teachers align such technologies with national curricular requirements. This study investigates the didactic potential of 3D pen activities designed by pre-service primary education teachers within the framework of the newly implemented National Curriculum (2023). The research explores how practical experiences with 3D pen integration manifest in the didactic reasoning of future teachers and how these activities align with selected curricular goals and performance standards. A qualitative content analysis was conducted on ten collaborative student projects (N = 50 participants). Each activity was evaluated using a goal–standard evaluation matrix (G1–G8 × S1–S7) on a 0–4 scale. Aggregated results were visualized through Sankey mapping to capture the intensity and structure of curricular alignment. The findings reveal strong alignment in areas related to project implementation, creativity, functionality, and problem-solving. Certain standards function as integrative nodes connecting multiple goals simultaneously. Less intensive representation was observed in communicative and regulatory dimensions. The study identifies a structured pattern of STEAM-oriented didactic reasoning, characterized by strong emphasis on project-based enactment and creativity, while communicative and regulatory dimensions remain comparatively underrepresented.
The advent of Industry 5.0 represents a paradigmatic shift toward human-centric manufacturing, where psychophysiological resilience of employees during collaboration with automated systems is crucial. While virtual reality (VR) has become standard for technical training, its potential for objective measurement of cognitive load using multimodal sensors remains a subject of research. This systematic review analyzes and synthesizes results from 57 experimental studies published in Web of Science, and Scopus databases between 2020 and 2026. Comparative analysis shows that while unimodal methods achieve average stress detection accuracy of 84.2
This narrative literature review presents a comprehensive synthesis of current knowledge in the field of educational robotics, focusing on unmanned aerial vehicles (UAV) and their impact on the cognitive architecture of the human brain. Based on a defined search strategy and inclusion criteria across relevant academic databases, this review examines the documented critical lack of spatial literacy in the population within the context of the emerging fourth and fifth industrial revolutions. Furthermore, it synthesizes existing literature proposing the implementation of drones as a key tool for neurocognitive development. Rather than providing new empirical data, this paper systematically reviews the mechanisms of manual piloting, FPV navigation, and drone programming, theoretically aligning these activities with Uttal’s taxonomy of spatial abilities. Finally, it consolidates previous research on pilots’ psychophysiological responses, including the phenomenon of cognitive dissonance during control inversion, to discuss broader implications for STEM education.
The rapid implementation of artificial intelligence (AI) tools in higher education is transforming learning processes, especially in the training of future teachers. Although AI is associated with personalized learning, greater independence, and academic effectiveness, empirical research on students’ attitudes in Lithuania and Romania remains limited. The study is based on a multidimensional concept of AI quality and aims to fill this gap by analysing how students evaluate AI across ergonomic, hedonic, and pragmatic dimensions and whether these evaluations differ across countries. A structured questionnaire with a 5-point Likert scale, adapted from previous studies, was used. The sample consisted of 188 prospective teachers: 93 in Lithuania (Vilnius and Klaipėda universities) and 95 in Romania (Valahia University). The data were analysed using confirmatory factor analysis and a second-order factor model (LISREL 9.3) to assess convergent validity and model fit. Both countries show positive assessments of AI in all dimensions. In Lithuania, a pragmatic and ergonomic orientation (functionality, learning benefits) prevails, while in Romania, a hedonistic orientation (attractiveness, motivation) prevails. The results show the need to differentiate AI integration strategies by national context and to strengthen both the pedagogical and motivational value of AI in pre-service teacher training.
This article investigates the use of automatic speech recognition (ASR) for voice-controlled interaction in computer games. It presents the design and implementation of two simple games (Tic-Tac-Toe and 8-Puzzle) developed in Unity and controlled using voice commands via the Whisper model. The study combines quantitative evaluation of recognition accuracy with qualitative analysis of user experience based on semi-structured interviews with 20 participants. The results show that modern ASR systems can achieve sufficient accuracy (approximately 84
The integration of digital technologies into education creates new opportunities for implementing student-oriented and inquiry-based approaches to learning. This article focuses on the application of inquiry-based learning in lower secondary education through the use of Scavenger Hunt Applications to create Interactive Educational Trails. Inquiry-Based Learning promotes active students’ engagement by positioning them as researchers and emphasizing observation, data collection, and reflection. The study presents a conceptual and practical framework for implementing all four levels of Inquiry-Based Learning, by using Scavenger Hunt Applications. Special attention is paid to the highest level of Inquiry-Based Learning, where students independently design and create interactive educational trails. The methodological section describes a case study with 7th grade students who developed an interactive educational trail using the AVAtrails application as part of the Erasmus+ AVATAR project. The results indicate that this approach increases students’ motivation, collaboration, digital skills, as well as their understanding of the surrounding environment. Despite the challenges students encountered during the creation of the Interactive Educational Trails, the results suggest that the combination of Inquiry-Based Learning, Scavenger Hunt Applications, and outdoor education represents an educational strategy with great potential for formal education.
Modern Smart Cities generate vast amounts of heterogeneous data from urban infrastructure, IoT ecosystems, cybersecurity events, and AI analytics. Existing solutions often treat these domains separately, resulting in fragmented knowledge that limits cross-domain reasoning and context-aware cybersecurity decision-making. Current ontologies are domain-specific, and knowledge graphs typically lack semantic integration and AI-supported reasoning. This paper presents an ontology-driven knowledge graph framework that unifies Smart City, cybersecurity, and AI-related data into a coherent semantic model. Organized into four layers—Data Acquisition, Information Structuring, Semantic Modeling, and Knowledge Reasoning—the framework transforms raw data into actionable, interpretable knowledge. By explicitly modeling entities, relationships, constraints, and dependencies, it supports context-aware reasoning, explainability, interoperability, and modular extensibility. Remaining challenges include scalable reasoning over large graphs, automated alignment of heterogeneous data, and continuous ontology evolution to accommodate changing urban systems and AI models. Addressing these challenges will strengthen the framework and advance ontology-driven, adaptive, and explainable cybersecurity governance in complex Smart City environments.
This study investigates the application of Topological Data Analysis (TDA) as a quantitative framework for detecting structural regime shifts in financial markets. Using intraday stock market data from 2016 to 2025, grouped by economic sectors, we transform price series into logarithmic returns and construct rolling daily time windows. For each window, persistent homology is computed on the corresponding point clouds, and temporal structural changes are quantified using the Wasserstein distance between consecutive persistence diagrams. The resulting time series of Wasserstein distances is compared with sectoral price dynamics to evaluate the correspondence between topological transitions and major market events. The analysis reveals that significant financial disruptions – most notably the 2020 global market crash associated with the COVID-19 pandemic – are consistently characterized by sharp increases in Wasserstein distance across all sectors. Additional sector-specific structural changes are detected during periods of macroeconomic instability, including the 2015–2016 selloff, the 2018 cryptocurrency crash, the 2022 geopolitical crisis, and subsequent regional market downturns. The findings suggest that persistent homology captures multiscale structural reorganizations in financial return distributions and provides a robust, model-independent indicator of market regime shifts. By focusing on geometric and topological invariants rather than solely on statistical moments or volatility measures, TDA offers a complementary perspective for financial risk assessment and systemic instability detection. Future research may further integrate topological descriptors with predictive modeling frameworks to enhance early-warning capabilities in financial markets.
Creativity and innovation skills have increasingly been recognized as key skills for the development of employability and lifelong learning. In the field of vocational technical education, these skills have specific relevance regarding the relationship with future professional performance. In the present research, the question is approached regarding the level to which students in a technical vocational school in Hungary perceive the support for the development of creativity and innovation skills in the school environment and the relationship with demographic and lifestyle-related variables. A quantitative methodological approach was chosen for the research. In the research, a total of 184 students completed the questionnaire. In the analysis, the relationships between the variables were examined with the help of Spearman rank-order correlation and linear regression analyses. In the research, it was found that the level of perceived support for the development of creativity and innovation skills is moderate or low. A weak negative relationship was found between the five-item index that indicates the level of support towards building creative and innovative abilities and the demographic factor of age. Moreover, a weak negative correlation was found between the index and perceived control over smartphone use, whereas a minor positive association was found between the index and sleep duration.
This paper introduces a gamified learning environment where human traders move beyond passive observation to compete directly against deep learning agents. Using simulations based on market data from 2025, human traders test their intuition against Multilayer Perceptron (MLP) models in direct competition for trading performance. Our findings highlight a distinct performance divergence: while deep learning agents capitalize on systematic patterns for high profitability, human traders demonstrate greater adaptability during market reversals and structural price shifts. By exposing behavioral biases like loss aversion through competitive benchmarking, this approach provides a rigorous, repeatable framework for teaching algorithmic market dynamics and risk management in finance.
We explore the application of the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm to portfolio optimization. Financial markets present high-dimensional and non-stationary environments where traditional reinforcement learning (RL) methods often suffer from value overestimation and instability. Unlike other approaches that use OHLCV (Open, High, Low, Close, Volume) data, we evaluate our approach using only a diverse set of normalized technical indicators. Such indicators are derived from OHLCV data; however, they are widely used by investors who rely on technical analysis. Our experimental results demonstrate that the agent achieves stable convergence. We further analyze the policy’s behavior using the Herfindahl–Hirschman Index (HHI) and portfolio turnover, revealing a shift toward more concentrated capital allocation during the inference phase.
The expansion of technology-enhanced learning environments has made digital interfaces central mediators of cognitive activity, positioning interface design as a determinant of learning effectiveness. While cognitive load theory is widely applied to instructional design, interface-induced cognitive load remains insufficiently systematised. This study proposes an interface-based framework for cognitive load assessment in digital learning contexts. The framework integrates cognitive load theory, human–computer interaction, and educational analytics, conceptualising cognitive load as a multidimensional construct shaped by visual, navigational, and interactional features. Subjective, behavioural, and analytical indicators are consolidated into a normalised, weighted integrated cognitive load index. By holding instructional content constant across interface configurations, the model isolates interface-induced extraneous load. Results indicate that variations in visual density, navigation complexity, and interaction sequencing significantly affect cognitive load, task efficiency, and cognitive fatigue. The study advances interface-centred cognitive load assessment and provides a scalable tool for comparative analysis and evidence-based design of inclusive digital learning environments.
This study aims to explore the use of Relief-F algorithm to screen key important car attributes from big data on consumer purchasing decisions. This car attribute can be used to create a questionnaire for distribution and then conduct statistical stepwise regression analysis on the collected questionnaire to obtain important attributes that meet the real needs of domestic consumers when purchasing cars. The research results showed that there were a total of 34 features from two car databases. After Relief-F feature filtering and stepwise regression screening, nine car attributes that consumers considered the most important were finally extracted. Further use of ANOVA to test the relationship between customers’ basic characteristics and important purchasing factors can be used to analyze the segmentation of the automotive market. With a clear definition of market segmentation, each automobile manufacturer can utilize resources to analyze their own market, while also defining direct competitors, potential competitors, and substitute competitors, forming the basis for SWOT analysis.
With the continuous expansion of e-commerce in China, emerging urban logistics applications need to address a wide range of challenges, including increasing demands and time-sensitive requirements. Seeking the balance among various factors such as transportation time, e-commerce enterprises enhance their market competitiveness by improving transportation service punctuality with reasonable operating costs. Based on the above background, this chapter first analyzes the urban logistics network structure and its related business process for the integrated warehouse and distribution mode and describes space-time trajectories of commodities and vehicles in the city logistics network. To eliminate existing bottlenecks, an optimization model of logistics service network is constructed through space-time modeling framework. Finally, a real-world instance is specifically tested based on Beijing logistics network data. The result shows that the optimization model can improve the service level of enterprises under the condition of limited resources and a large number of orders.
Aiming at the problems of complex entity relationship, low pipeline efficiency, and insufficient model accuracy in the construction of knowledge graph in walnut pest and disease field, this chapter proposes a new knowledge graph construction in walnut pest and disease field based on deep learning. By adopting Begin-Inside-Outside-End-Single (BIOES) annotation, the efficiency of knowledge graph ternary extraction is effectively improved, and at the same time, iterated dilated convolutional neural network (IDCNN) is introduced to improve the BERT-CRF model. The introduction of dilated convolution extends the network sensory range without parameters, improves the modeling ability of long-distance dependencies, and then enhances the comprehension and accuracy of the model on the contextual information of text sequences. The experimental results show that BERT-IDCNN-CRF achieves an excellent F1 score of 73
Due to the complexity of the iron ore sintering process and the harsh on-site production environment, the industrial data collected often exhibit strong coupling, high noise, and significant time delays. Existing models find it difficult to deeply learn the complex relationships within sintering data. To improve the performance of the sintering endpoint prediction model and reduce the impact of complex relational patterns in the data, this chapter proposes an encoder-decoder model based on graph neural networks and attention mechanisms. First, a graph adjacency matrix is constructed driven by the sintering data to efficiently model the complex relationships between sintering parameters. Next, a factorized-temporal attention module is designed, introducing predictive results to quantify the importance of variables and time steps, enabling the model to accurately focus on key information while reducing the impact of redundant information. Finally, a long short-term memory (LSTM) network is used for temporal feature extraction, thoroughly capturing latent patterns in the sequence, resulting in more accurate prediction outcomes. Experimental results for sintering endpoint prediction show that the proposed model reduces mean absolute error and mean square error by 32.76
In the increasingly thriving digitalization external environment, sustaining competitive advantage in supply chains poses significant challenges for companies. Effective management of supplier and customer concentration is essential for maintaining this advantage during digital transformation. This study investigates the relationships between supplier concentration, customer concentration, and digital transformation of firms in the Chinese market context. Based on transaction cost theory and resource dependence theory, this study assumes that highly concentrated suppliers will hinder enterprises’ digital transformation. Grounded in bargaining power and operation management, it is proposed that successful digital transformation can reduce enterprises’ dependence on customers. Utilizing data from Chinese-listed A-share companies from 2010 to 2022, the study reveals the firms that may progress more slowly. Conversely, digital transformation effectively diversifies customer bases, strengthens customer relationships, and enhances risk management through a more resilient business model. The results provide important insights on how to enhance a company’s market competitiveness and innovation by reducing supplier concentration and implementing effective digital transformation.
