Abstract
The Internet of Things generates massive data streams, with edge computing emerging as a key enabler for online IoT applications and 5G networks. Edge solutions facilitate real-time machine learning inference, but also require continuous adaptation to concept drifts. While extensions of the Very Fast Decision Tree (VFDT) remain state-of-the-art for tabular stream mining, their unregulated growth limit efficiency, particularly in ensemble settings where post-pruning at the individual tree level is seldom applied. This paper presents DFDT, a novel memory-constrained algorithm for online learning. DFDT employs activity-aware pre-pruning, dynamically adjusting splitting criteria based on leaf node activity: low-activity nodes are deactivated to conserve resources, moderately active nodes split under stricter conditions, and highly active nodes leverage a skipping mechanism for accelerated growth. Additionally, adaptive grace periods and tie thresholds allow DFDT to modulate splitting decisions based on observed data variability, enhancing the accu-racy–memory–runtime trade-off while minimizing the need for hyperparameter tuning. An ablation study reveals three DFDT variants suited to different resource profiles. Fully compatible with existing ensemble frameworks, DFDT provides a drop-in alternative to standard VFDT-based learners. © 2026, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.

Abstract
This study presents a photovoltaic (PV)-based electric vehicle (EV) charging system designed to optimize energy use and support isolated microgrid operations. The system integrates PV panels, DC/AC, AC/DC, and DC/DC converters, voltage and frequency droop control, and two energy management algorithms: Power Sharing and SEWP (Spread Energy with Priority). The DC/AC converter demonstrated high efficiency, with stable AC output and Total Harmonic Distortion (THD) limited to 1%. The MPPT algorithm ensured optimal energy extraction under both gradual and abrupt irradiance variations. The DC/DC converter operated in constant current mode followed by constant voltage regulation, enabling stable power delivery and preserving battery integrity. The Power Sharing algorithm, which distributes PV energy equally, favored vehicles with a higher initial state of charge (SOC), while leaving low-SOC vehicles at modest levels, reducing satisfaction under limited irradiance. In contrast, SEWP prioritized low-SOC EVs, enabling them to achieve higher SOC values compared to the Power Sharing algorithm, reducing SOC dispersion and enhancing fairness. The integration of voltage and frequency droop controls allowed the station to support microgrid stability by limiting reactive power injection to 30% of apparent power and adjusting charging current in response to frequency deviation.

Abstract
In light of growing concerns about environmental degradation and unsustainable consumption patterns, effective communication about sustainability has become crucial in encouraging more environmentally conscious behavior. With their ability to connect with large audiences, social media influencers hold considerable potential to foster environmental awareness and promote sustainable lifestyles. This study explores how influencers communicate sustainability and how they can produce content that maximizes audience engagement. First, a systematic literature review identifies key research gaps and synthesizes relevant insights to inform the analysis of influencer-driven sustainability discourse. Building on this foundation, the study applies a two-phase, data-driven approach. Using a dataset of 2858 Instagram posts from 25 green influencers, we apply Latent Dirichlet Allocation (LDA) to identify prevalent themes published by green influencers and to track their evolution over time. The second phase involves developing a multimodal predictive model that estimates engagement levels by integrating textual and visual embeddings with structured posting metadata. The findings have direct implications for content strategy: lifestyle and fashion-oriented sustainability themes consistently attracted higher engagement, while technical subjects such as gardening and nutrition received less interaction. Over time, influencers' content exhibited clearer topic delineation, indicating a maturation of sustainability narratives. Furthermore, caption content was found to be the strongest predictor of engagement, though visual and contextual features also added value. This study offers a practical framework for enhancing engagement through targeted, data-informed content strategies. These insights can inform the work of influencers, brands and campaign designers looking to maximize the impact of environmentally focused messaging.

Abstract
Malicious activity detection in honeypot environments remains challenging due to the volume and heterogeneity of captured data, as well as the sequential nature of attacker behavior. This study proposes an integrated framework combining a Cowrie-based honeypot with a locally deployed Large Language Model (LLM) to enable automated detection and near-real-time behavioral analysis. Attacker interactions are captured and processed through a structured preprocessing stage that reconstructs session-level activity. These representations are analyzed using an LLM, allowing contextual interpretation of authentication patterns, command execution sequences, and post-compromise behavior. Structured analytical outputs are generated, including severity classification, reasoning, and recommended actions. Evaluation was conducted using isolated and concurrent attack scenarios. Results indicate effective identification of brute-force attacks, reconnaissance activity, persistence staging, and download-and-execute patterns, achieving a 94.23% Accuracy (95% CI: 84.05–98.79%), 100.0% Precision, 87.50% Recall, and an F1-score of 93.33% across an expanded evaluation of 52 independent observations, with zero false positives (FPs). These figures are derived from a single evaluation run and are reported as preliminary, proof-of-concept estimates rather than as stable, production-grade performance. Analytical output remained stable for sessions governed by deterministic severity overrides. However, a low-intensity multi-stage session produced inconsistent severity classifications under concurrent conditions, indicating that analyst review is still required for borderline cases. Integration of deterministic preprocessing with LLM-based reasoning enables the transformation of raw honeypot logs into structured and actionable cybersecurity intelligence, reducing analyst workload while improving the explainability and reliability of intrusion analysis in near-real-time environments.

Abstract
Purpose The primary objective of this study is to investigate the role of entrepreneurial team stability in the formation, consolidation, and evolution of collective entrepreneurial identity within the context of incubation programs. Specifically, the research aims to examine how long-term relationships, complementary skill sets, shared vision and adaptive capabilities contribute to team cohesion and resilience, and how these factors collectively influence the development of a collective entrepreneurial identity. Design/methodology/approach It examines five entrepreneurial teams that completed the CEU iLab Incubation Program in Budapest, each associated with a distinct startup across diverse sectors and industries. Data collection was predominantly retrospective, focusing on reconstructing and interpreting the past experiences of team members that contributed to their current entrepreneurial trajectories. Findings Findings underscore the importance of stability, cohesion and adaptability in shaping entrepreneurial team identity. Teams that cultivate long-term relationships, complementary skill sets and a shared vision demonstrate higher trust, psychological safety and mutual understanding, which facilitate collaboration, conflict resolution and effective decision-making. Open communication and adaptability to market changes further reinforce team cohesion and resilience. Originality/value The results suggest that stable, skill-diverse teams are better positioned to navigate uncertainties while maintaining a strong, collective entrepreneurial identity. These insights offer practical implications for incubators and policymakers, emphasizing the need to support team continuity alongside venture growth metrics to foster sustainable startup development.

Abstract
Recent technological advancements have increased computer crime, requiring public authorities to implement structured mitigation strategies. While initiatives exist to improve digital literacy on device security, they must also address the complexities of computer crime. Using Design Science Research, this study investigated the applicability of chatbots to raise awareness of computer crime in a public administration setting. A systematic literature review highlighted the issue's relevance and identified knowledge gaps. A scoping review gathered concepts, methodologies, technologies, architectures, and tools for developing and evaluating an effective chatbot. The design and development phase included a detailed proposal for a sophisticated chatbot architecture. During the demonstration and evaluation phases, the utility of the chatbot was tested in the domain of conversational flow efficiency and usability. The study's primary results and contributions are to assess the chatbot's effectiveness in raising awareness of computer crime on public websites. Future work should focus on implementing the chatbot in the actual context of public administration, proposing a network of specialized conversational assistants, and improving public service interoperability to enhance computer crime awareness.