ABSTRACT This study examines the evolution of artificial intelligence (AI) in auditing by systematically synthesizing 26 years of global scholarship to address conceptual fragmentation in the field. Guided by the PRISMA protocol, it analyses 269 Scopus-indexed journal articles, harmonized using OpenRefine and biblioMagika, and examined through performance analysis, co-occurrence mapping, and life-cycle modeling with biblioMagika, VOSviewer, and Biblioshiny. Integrating behavioral, technological, and governance perspectives, the study clarifies the domain’s intellectual foundations, thematic structure, and temporal development. Findings show rapid expansion since 2019, reflecting a shift from early rule-based and expert systems toward advanced machine learning, explainable AI, and generative AI applications. Six thematic clusters organize the literature: computational audit analytics; governance and responsible AI; AI–audit integration; blockchain-enabled audit evidence; automation and audit quality; and analytics-driven continuous auditing. Temporal evidence indicates an increasing normative orientation, with stronger emphasis on transparency, accountability, trust, and ethical governance. Life-cycle modeling suggests the field remains in a steep growth phase, indicating substantial scope for further theoretical and practical advancement. Although limited to Scopus-indexed journal articles, the results offer a foundation for research on auditor–AI interaction, behavioral effects of automation, governance mechanisms for AI assurance, and institutional variation in technology adoption. By providing a longitudinal science-mapping analysis, the study consolidates publication trends, intellectual structure, thematic evolution, and life-cycle forecasting, and identifies research opportunities likely to shape audit methodology, governance, and professional judgment in an era of intelligent systems.

Purpose This study aims to introduce and empirically validate a novel moderated-mediation model to explain how integrating AI and blockchain as an intelligent information system for accounting transforms organizational knowledge into a strategic asset, ultimately enhancing sustainable value and competitive advantage within the ASEAN region. The authors propose a comprehensive framework where this technological system improves sustainability reporting quality (SRQ; a critical knowledge output), contingent upon the firm’s internal governance and the external environment. Specifically, the authors pioneer a contingent governance perspective, arguing that the amplifying role of internal governance is itself conditional on external pressures. Design/methodology/approach Adopting a quantitative approach, this study analyzes a panel data set of 3,300 firm-year observations from listed companies across major ASEAN economies. Structural equation modeling is used to test a sophisticated moderated-mediation framework. This methodology allows for a robust examination of the complex, conditional pathways linking AI and blockchain adoption to sustainable value while simultaneously assessing the contingent effects of enterprise risk management (ERM) and environmental uncertainty. Findings Results reveal that the AI and blockchain-powered system significantly improves the quality of sustainability reporting (a key knowledge output), which in turn drives sustainable value. Critically, the effectiveness of this system is amplified by robust ERM. However, the study uncovers a nuanced conditional effect: the strategic value of ERM is itself contingent on the level of external environmental uncertainty, highlighting how internal knowledge systems must adapt to external pressures to remain effective. Practical implications This research provides a practical roadmap for managers on how to build and govern information systems that translate technological investment into tangible, sustainable performance. It offers a new lens for investors to evaluate a firm’s knowledge management capabilities alongside its technological infrastructure. For policymakers, it underscores the need for policies that foster not just technology adoption but also the development of resilient organizational systems capable of navigating a highly interdependent world. Originality/value First, it is among the first study, to the best of the authors’ knowledge, to empirically model the complete pathway from a specific technological system to sustainable value via the mediating mechanism of SRQ. Second, and most significantly, it pioneers the concept of “contingent governance” by demonstrating that the moderating effect of ERM is not universal but is itself moderated by environmental uncertainty. This multi-level interaction offers a far more nuanced view than previously understood. Third, by situating this complex model within the under-researched, high-growth context of the ASEAN region, the authors provide crucial insights for emerging markets on leveraging information as a predominant asset.

A Accurate forecasting of household energy consumption is critical component of smart home energy management systems, enabling efficient energy utilization, cost reduction, and demand-side planning. This work presents a deep learning–based framework for short-term energy consumption forecasting using real-world household power usage data. The proposed approach integrates one-dimensional Convolutional Neural Networks (CNN) with sequential learning models—Long Short-Term Memory (LSTM), Bidirectional LSTM (BiLSTM), and Gated Recurrent Units (GRU)—to capture both local temporal patterns and long-term dependencies in time-series energy data. The dataset is preprocessed through hourly resampling and Min–Max normalization, followed by sequence generation using a 24-hour sliding window. Model performance is evaluated using Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) on unseen test data. Comparative analysis demonstrates that hybrid CNN-based architectures outperform standalone temporal models by effectively learning spatial–temporal features. Among the evaluated models, the best-performing architecture achieves the lowest RMSE and MAE, indicating strong predictive capability. The results confirm the suitability of CNN RNN hybrid models for smart home energy forecasting and provide a scalable foundation for future intelligent energy management systems. Key Words: Smart Home Energy Management, Energy Consumption Forecasting, Time Series Prediction, Deep Learning, Convolutional Neural Network (CNN), LSTM, BiLSTM, GRU, Hybrid Models, Real-Time Energy Data, RMSE, MAE

Cognitive Aware Temporal Learning for Hyper-Personalized Education develops intelligent education systems which transform conventional teaching methods into automatically adjusting educational experiences. The AI-based personal tutoring system functions as more than a basic question and answer platform by storing study session details from past sessions to track student development throughout their academic journey. The system creates knowledge timelines to discover user learning weaknesses which it addresses through spaced repetition methods and custom quizzes that adjust based on each user's unique study behavior. The platform uses advanced Natural Language Processing and LLaMA-2/3 (7B) with Sentence-BERT/MiniLM embedding models to provide temporal reasoning and vector storage through FAISS and session tracking with SQLite/PostgreSQL database systems. The system demonstrates knowledge advancement through time which supports learning by adjusting content based on intelligent algorithms. The Python based system uses Streamlit/Flask to create an interactive interface which enables this "temporal tutor" to provide customized educational assistance as students progress through their educational paths in coding and exam preparation. Keywords:AI-powered personal tutor, Temporal memory, Adaptive learning, Spaced Repetition, Knowledge Timelines, Personalized Education, Session Tracking, Temporal Reasoning, Embeddings.

Rapid advancements in foundation models, including Large Language Models, Vision-Language Models, Multimodal Large Language Models, and Vision-Language-Action models, have opened new avenues for embodied AI in mobile service robotics. By combining foundation models with the principles of embodied AI, where intelligent systems perceive, reason, and act through physical interaction, mobile service robots can achieve more flexible understanding, adaptive behavior, and robust task execution in dynamic real-world environments. Despite this progress, embodied AI for mobile service robots continues to face fundamental challenges related to the translation of natural language instructions into executable robot actions, multimodal perception in human-centered environments, uncertainty estimation for safe decision-making, and computational constraints for real-time onboard deployment. In this paper, we present the first systematic review of foundation models in mobile service robotics, following the preferred reporting items for systematic reviews and meta-analysis (PRISMA) guidelines. Using an OpenAlex literature search, we considered 7506 papers for the years spanning 1968–2025. Our detailed analysis identified four main challenges and how recent advances in foundation models, related to the translation of natural language instructions into executable robot actions, multimodal perception in human-centered environments, uncertainty estimation for safe decision-making, and computational constraints for real-time onboard deployment, have addressed these challenges. We further examine real-world applications in domestic assistance, healthcare, and service automation, highlighting how foundation models enable context-aware, socially responsive, and generalizable robot behaviors. Beyond technical considerations, we discuss ethical, societal, human-interaction, and physical design and ergonomic implications associated with deploying foundation-model-enabled service robots in human environments. Finally, we outline future research directions emphasizing reliability and lifelong adaptation, privacy-aware and resource-constrained deployment, as well as the governance and human-in-the-loop frameworks required for safe, scalable, and trustworthy mobile service robotics.

The application of artificial intelligence algorithms in news media systems is gradually forming an intelligent content ecosystem driven by deep models. We have constructed a multi module intelligent system that integrates Transformer, BERT, and knowledge graph, covering three major functions: semantic parsing, content recommendation, and intelligent generation. Based on multimodal embedding and Prompt optimization techniques, we have completed algorithm iterations on the basis of traditional CNN and LSTM architectures. The experiment showed that the BART model improved the BLEU score to 36.42 in the abstract generation task, which was 16.4% higher than that of Transformer; The Transformer in the recommendation system NDCG@10 Reaching 0.873, better than XGBoost and LSTM. The system exhibits significant advantages in both accuracy and response speed.

Discharge follow-up refers to the method of observation and treatment of patients treated in the hospital after discharge, in which medical workers continue to pay attention to the changes of patients’ condition and rehabilitation, and continue to guide patients’ rehabilitation. Encourage the continuous improvement of hospital service quality and patient satisfaction. An optimized design scheme for an intelligent patient tracking system based on the Internet of Things and sensor technology has been proposed. The Kalman filtering algorithm in sensor technology is applied to improve the patient tracking problem of the IoT intelligent tracking system constructed in this paper, thereby expanding the scope and accuracy of the system. Finally, the simulation test and analysis are carried out. The simulation results show that, in the simulated follow-up scenarios for neurology patients (including home rehabilitation environments and multi-target intersection scenes similar to community activity areas), the algorithm is 10.22% more accurate than the traditional algorithm.The simulation results show that the algorithm is 10.22% more accurate than the traditional algorithm. The results show that the intelligent follow-up system is reasonable and practical, and basically meets the requirements of the follow-up management of specialized cases in the Neurology Department of our hospital. The development and research of intelligent follow-up technology for neurology patients based on sensor technology has important practical significance for guiding the follow-up rehabilitation treatment of neurology diseases and patients.

Recent advances in digital infrastructures, cloud services, blockchain platforms, Internet of Things (IoT), and artificial intelligence (AI) have significantly increased the complexity of cybersecurity threats while intensifying concerns related to data privacy and system trust. In response, contemporary research has focused on integrating intelligent data-driven techniques with secure and privacy-aware mechanisms to counter sophisticated cyberattacks. This literature review systematically analyzes and synthesizes selected peer-reviewed studies with an emphasis on AI-driven intrusion detection systems and blockchain-enabled security architectures. The reviewed works are examined in terms of their underlying methodologies, algorithms, tools, strengths, and limitations. A comparative analysis identifies key trends, persistent challenges, and research gaps, particularly in explainable AI, system scalability, and secure analytics. The findings highlight the need for unified and deployable cybersecurity frameworks that balance detection accuracy, transparency, and operational efficiency. This review provides a structured foundation to support future research on intelligent and trustworthy cybersecurity systems.

In the context of the accelerating digital transformation of higher education worldwide, Artificial Intelligence (AI) has emerged as a pivotal driver for innovating online learning models, enhancing teaching quality, and optimizing learners’ experiences. In Vietnam, private universities - characterized by institutional flexibility, intense competitive pressure, and a high degree of financial and academic autonomy - are increasingly required to adopt AI as a strategic instrument to improve educational quality and institutional competitiveness. Despite its significant potential, the implementation of AI in online training across Vietnamese private universities remains fragmented, unsystematic, and constrained by multiple challenges related to technological infrastructure, human resources, financial capacity, governance mechanisms, as well as ethical and legal frameworks. This study aims to provide a comprehensive examination of the current state of AI application in online education at private universities in Vietnam, to identify key barriers hindering effective implementation, and to propose strategic solutions tailored to the national and institutional context. Employing a mixed-methods research design, the study integrates quantitative survey data with qualitative insights derived from in-depth interviews with university leaders, faculty members, and students. The findings indicate that AI has been preliminarily applied in areas such as intelligent learning management systems, personalized learning pathways, AI-powered chatbots, and learning analytics. However, the depth and effectiveness of these applications remain limited, and their transformative potential has yet to be fully realized. Drawing on empirical evidence and international best practices, the paper proposes a strategic framework encompassing institutional governance reform, digital and AI-oriented capacity building, investment in data-driven infrastructure, and the development of a sustainable AI-enabled educational ecosystem for private higher education institutions. This study contributes to the growing body of literature on AI in higher education within developing country contexts, while also offering practical managerial and policy implications for advancing the digital transformation and long-term sustainability of Vietnam’s private university sector

This study aims to analyze the degree of air superiority, a strategic concept thatplays a central role in the planning and implementation of modern air operations. This termrefers to the degree of control of an air force over a specific area that allows the execution ofmissions without significant interference from opposing air forces. In general, the degree of airsuperiority is classified into three categories: air parity, air superiority, and air supremacy, eachof which indicates a different degree of dominance over potential enemy air threats. Airdominance is a strategic prerequisite that determines the outcome of modern conflicts. Theconcept of degree of superiority evolves along with changes in technology, doctrine, and thecomplexity of global geopolitics. In the context of air operations, achieving degree of airsuperiority is influenced by various factors, including technological capabilities, theeffectiveness of air defense systems, information mastery, and the efficiency of command,control, communications, computers, intelligence, reconnaissance, and surveillance (C4ISR)systems. The interaction of these factors contributes to the success of joint operations and theability to maintain long-term air dominance. Effective air superiority also directly impacts thefreedom of action of one's own forces in conducting cross-service operations. Acomprehensive understanding of the degree of air superiority is a key factor in strategicdecision-making at the operational level. Integration of technological capabilities, informationsuperiority, and cross-service coordination is necessary to achieve sustainable air dominance.Therefore, analyzing the degree of air superiority is not only crucial for tactical purposes butalso serves as the basis for developing comprehensive air defense doctrine and strategy. Thisarticle discusses the theoretical framework in depth, the factors influencing its achievement,and presents case studies from various conflicts ranging from World War II to potential conflictsin the Indo-Pacific. This study confirms that successfully achieving air superiority depends notonly on air power alone, but also on multi-domain integration, defense industry readiness, andpolitical and diplomatic support.

ABSTRACT Personalized Learning through Cognitive State Analysis is an intelligent web-based learning system that monitors a learner’s attention level using real-time webcam analysis. The system detects facial presence and eye activity using OpenCV and Haar Cascade algorithms to determine whether the learner is attentive or distracted. Based on the detected cognitive state, the system dynamically adapts the learning content. If the learner is attentive, video-based learning continues. If the learner is distracted multiple times, the system recommends alternative learning modes such as text-based content or quizzes. The platform also provides a dashboard to analyze attentive time, distracted time, and overall performance. This system aims to improve learning efficiency, engagement, and personalization in digital education environments. Keywords: Personalized Learning, Cognitive State Analysis, Haar Cascade, OpenCV, Adaptive Learning, Attention Detection, E-learning.

Motion and path planning are fundamental to intelligent robotic systems, enabling navigation. The objective is to generate collision-free trajectories in obstacle-rich configuration spaces (C-spaces) while meeting performance constraints. In environments with narrow passages planning becomes especially difficult, as feasible regions have low measure and are rarely reached by random sampling. Classical sampling-based planners are probabilistically complete but inefficient in such regions. Learning-based planners like MPNet offer fast inference but often produce infeasible paths in cluttered areas, requiring expensive postprocessing. To address this trade-off, we propose a hybrid framework that combines improved sampling, structural abstraction, and neural prediction. A modified bridge-test sampler applies directional perturbations and corridor checks to generate reliable narrow passage samples. These are clustered into a sparse set of representative bridge points, which serve as nodes in a global graph. At query time, a greedy heuristic search explores this graph, using a neural local segment generator to connect nodes. We validate the approach on 2D maze maps, 3D voxel environments, and a 12-DOF manipulator performing a plugging task inside a simulated nuclear steam generator. Across all tasks, our method significantly outperforms classical and learning-based baselines in terms of success rate and planning time in narrow-passage-dominated scenarios. The inclusion of the repair module, under relaxed assumptions, also allows the framework to retain a generalized form of probabilistic completeness.

Abstract 3D object detection plays a vital role in autonomous driving and intelligent measurement systems. Transformer-based frameworks have recently gained prominence by predicting sampling points and directly extracting features from the image space. This design bypasses explicit depth estimation, Bird's Eye View (BEV) projection, and post-processing steps such as Non-Maximum Suppression (NMS), thereby significantly improving inference speed. However, these methods often lag behind view transformation-based approaches in detection accuracy, particularly in the quantitative estimation of object size, orientation, and motion parameters. This paper attributes this performance gap to the difficulty of extracting reliable object features from the sampled locations. Since the sampling points are only indirectly supervised by detection losses, many of them often do not accurately correspond to actual object regions, resulting in the collection of background or irrelevant features. Furthermore, existing motion-based strategies for aligning sampling points across temporal frames often suffer from significant errors, which further compromise the quality of aggregated temporal features. To tackle these challenges, we introduce a novel 3D object detection and measurement framework that integrates two core components: Point-Driven Sampling, which leverages supervision from point clouds to direct the sampling process toward object-relevant regions during training, thereby enhancing the fidelity of geometric measurements; and Center-Aligned Temporal Aggregation, which uses a learnable module to adaptively align and fuse temporal features, eliminating the need for explicit motion compensation while improving dynamic measurement accuracy. Experiments on the nuScenes benchmark validate the effectiveness of our approach, demonstrating a favorable trade-off between measurement accuracy and inference speed.

The transformation of financial management in the digital era extends beyond automation and software integration; it represents a structural redefinition of how organizations interpret, institutionalize, and leverage data for strategic advantage. While firms increasingly invest in artificial intelligence (AI), predictive analytics, and blockchain-enabled systems, empirical evidence suggests that technological adoption alone does not guarantee superior financial outcomes. This study develops and validates a multidimensional structural model examining how Digital Adoption Level (DAL), Data Intelligence Capability (DIC), and Organizational Mindset Orientation (OMO) collectively influence Strategic Financial Performance (SFP). Using a descriptive-analytical research design supported by secondary data synthesis (2020–2025), thematic coding of 52 peer-reviewed studies, and quantitative modeling through multiple regression and Structural Equation Modeling (SEM), this study tests mediation pathways and structural alignment mechanisms. Results demonstrate that Data Intelligence Capability exerts the strongest standardized effect (β = .62, p < .001), followed by Organizational Mindset Orientation (β = .55, p < .001), while Digital Adoption Level exhibits a conditional impact (β = .41, p < .01). Bootstrapped mediation analysis confirms partial mediation of leadership cognition between digital adoption and performance. Findings suggest that financial transformation is intelligence-centric rather than technology-centric. Sustainable performance emerges when digital infrastructure is cognitively interpreted and structurally embedded within governance systems.

Abstract Training lightweight traffic object detectors with knowledge distillation (KD) is crucial for intelligent transportation systems under resource-limited conditions. However, most KD approaches still rely on fixed thresholding to generate binary masks for student feature reconstruction, disrupting semantic continuity in complex traffic scenes. In this work, a Dual Attention Distillation Framework (DADF) using Adaptive Feature Fusion is proposed for traffic object recognition. Instead of binary masks, DADF produces Softmax-based normalized distributions soft masks along both spatial and channel dimensions, thereby more effectively regulating the continuity of semantic information. To adaptively balance spatial and channel cues, teacher feature variances are utilized for weighting and fusing the masks into a unified attention map. Meanwhile, a multilayer perceptron (MLP) generator is subsequently used to reconstruct the masked student features. Finally, the distillation process is optimized by minimizing the mean squared error (MSE) between the reconstructed and teacher features. We extensively validated the effectiveness of the DADF method across multiple datasets and detectors. On Cityscapes, it boosts YOLOv8 mAP from 41.9% to 44.1%, while cutting parameters and GFLOPs by 73.0% and 71.6%, and raising inference speed from 188.7 to 202.8 FPS. On KITTI, DADF boosts the RT-DETR mAP from 85.8% to 90.5%, even surpassing its teacher model. It also cuts parameters by 31.0%, reduces GFLOPs by 32.5%, and increases speed from 33.8 to 35.3 FPS. These results highlight DADF’s suitability for traffic measurement applications under resource constraints.

Purpose The Fourth Industrial Revolution has accelerated technological advancements across industries, necessitating that countries, research institutions and enterprises enhance their technological competitiveness. A key challenge in this process is the ability to predict promising technologies and integrate them into strategic decision-making. However, existing methods predominantly rely on expert-driven qualitative assessments, which can be subjective and inconsistent. This study aims to address these limitations by proposing a quantitative, data-driven framework for technology foresight and strategic development in Korea’s railway industry, with a specific focus on emerging digital and control systems. Design/methodology/approach This research integrates autoregressive integrated moving average (ARIMA) time-series forecasting and weighs social network analysis (SNA) to systematically identify emerging technological trends. Using 4,352 railway-related patents from the Korean Intellectual Property Office (KIPO) from 1990 to 2023, technology keywords were extracted through text mining using TF-IDF scores. Promising technologies were identified by analyzing their temporal growth patterns (including forecast confidence intervals) and network influence, enabling a data-driven approach to forecasting technological developments and informing strategic planning. Findings The analysis demonstrates that the synergistic use of ARIMA-based forecasting and SNA-driven influence assessment provides a robust and systematic methodology for identifying emerging technologies. The results highlight that core technologies related to “control,” “signal,” “sensor,” “device” and “speed” are poised for significant growth and hold central positions within the technology network. This quantitative approach enhances technology management by reducing reliance on subjective expert opinions and providing objective, data-driven insights. Practical implications This study offers a structured methodology for organizations to enhance technology foresight and strategic planning. By leveraging predictive analytics, policymakers and industry leaders can proactively identify high-potential technologies, optimize resource allocation and foster innovation in the railway sector, particularly in the transition toward automated and intelligent transportation systems. Originality/value This research contributes to the field of technology forecasting by introducing a reproducible, quantitative framework that combines time-series analysis with network theory. By justifying the methodological choices and demonstrating their synergy, this framework offers a novel and robust alternative to traditional methods for strategic decision-making and technology development, particularly in mature, high-tech industries like the railway sector.

Autonomous driving technology is a core enabler for new energy vehicle industrial upgrading and a critical pillar for achieving sustainable development goals (SDGs), especially sustainable urban mobility, low-carbon transportation, and efficient intelligent transportation systems (ITS). However, unstable nighttime low-light perception severely restricts autonomous driving deployment, hindering sustainable transportation development—rooted in visual feature degradation and cross-modal imbalance that impair 3D object detection (autonomous driving’s core perception technology). To address this and advance sustainable autonomous driving, this paper proposes a Bird’s-Eye View (BEV)-based multi-modal 3D object detection approach tailored for nighttime scenarios, integrating low-light adaptive components while preserving the original BEV pipeline. Without modifying core inference, it enhances low-light robustness and cross-modal fusion stability, ensuring reliable perception for sustainable autonomous driving operation. Extensive experiments on the nuScenes nighttime subset quantify performance via rigorous metrics (NDS, mAP, mATE). Results show the method outperforms BEVFusion with negligible parameter/inference overhead, achieving 1.13% NDS improvement. This validates its effectiveness and provides a sustainable technical tool for autonomous driving perception, promoting new energy vehicle popularization, optimizing urban ITS efficiency, reducing perception-related accidents and carbon emissions, and directly contributing to transportation and socio-economic sustainability.

Contemporary gastronomy is experiencing a structural transformation driven by technological innovation, cultural reconfiguration, sustainability, and digitalization. Rather than breaking with tradition, culinary evolution reflects a dynamic reinterpretation that integrates scientific knowledge, creative practices, and environmental responsibility while preserving cultural heritage. This narrative review analyzes interactions between innovation, tradition, sustainability, and digital transformation using peer-reviewed studies (2020–2025). Findings show that molecular gastronomy, artificial intelligence, and digital systems expand culinary possibilities, while local ingredients and ancestral knowledge reinforce identity. An integrative four-axis model is proposed, highlighting balance between authenticity and modernization as the sector’s key strategic challenge.

Early detection of diseases plays a crucial role in improving patient survival rates and reducing medical complications. However, limited access to healthcare professionals, cost barriers, and delayed medical consultations often prevent timely diagnosis. HealthAI is a web-based intelligent disease prediction system developed to provide preliminary healthcare guidance using Natural Language Processing (NLP) and Machine Learning techniques. The system processes user-entered symptoms using TF-IDF vectorization and cosine similarity algorithms to match symptoms against a dataset of 200 diseases. The platform integrates an AI chatbot interface, emergency symptom detection, voice input processing, and hospital location services. Experimental evaluation demonstrates prediction confidence scores ranging from 90% to 98%. HealthAI provides fast, accessible, and reliable healthcare assistance, enabling early-stage disease awareness and timely medical intervention..

Customer Lifetime Value (CLV) estimation over the observed transactional horizon is a fundamental challenge in hospitality analytics, supporting revenue management, personalization, and long-term customer relationship strategies. However, existing models predominantly rely on structured behavioral data while overlooking the emotional intelligence embedded in guest narratives. This study proposes an interpretable multimodal deep learning (DL) framework that bridges behavioral and emotional intelligence for CLV estimation by integrating structured booking records with unstructured hotel review text. Model interpretability is ensured through SHAP analysis for structured attributes, LIME for local textual explanations, and attention visualization for modality interaction analysis. Experimental evaluation on large-scale hospitality datasets demonstrates that the proposed multimodal framework outperforms traditional machine learning models, unimodal deep learning baselines, and classical ensemble learners, yielding consistent improvements across multiple error metrics and a notable increase in goodness of fit. The results confirm that emotional intelligence extracted from guest reviews significantly enhances CLV estimation and provides actionable insights for hospitality decision-making, supporting the deployment of transparent and explainable artificial intelligence (XAI) systems for strategic customer value management.