Introduction: Decentralized autonomous organizations (DAOs) are an emerging organizational form that operates entirely on blockchain infrastructure. Within a DAO, organizational governance rules are hardcoded in transparent and immutable smart contracts. In principle, these rules are intended to facilitate decentralized decision-making among token holders who collectively create, discuss, and vote on proposals that govern the organization. Despite their promise, the extent to which DAOs achieve true decentralization in practice remains unclear. This study addresses an underexplored area in the literature by empirically investigating key aspects of DAO governance, particularly the degree of decentralization and participant composition.Method: Network analysis is used to examine proposal voting coalitions among participants as a proxy for decentralization. Sentiment analysis is employed to assess trust among participants. The analysis draws on data from 54 DAOs, including 774 unique proposals and 13,085 associated token holder comments.Results: The findings indicate that DAOs may not achieve the level of decentralization originally envisioned. Moreover, decentralization and participant composition within governance structures play a critical role in shaping trust, voting participation, and overall financial performance in DAOs PracticalImplications: Although current voting mechanisms aim to reduce the dominance of large token holders (whales), DAOs may still fall short of the level of decentralization originally envisioned. Accordingly, more advanced voting mechanisms may be required to further mitigate coordination and strategic behavior in proposal voting. In addition, DAOs could benefit from adjusting the threshold requirements for the Foundation to improve accessibility for token holders and encourage broader participation. Leveraging artificial intelligence (AI) may also help streamline the voting process and improve the clarity of proposals.
PurposeGrounded in the dynamic capabilities view and organisational learning theory, this study examines the relative association of big data analytics-enabled dynamic capabilities (BDAEDC) on the quality of the decisions made in organisations through the mediating roles of exploitative and explorative learning and through the moderating role of top-management support for learning (TMSL).Design/Methodology/ApproachUsing cross-sectional survey data collected from 330 software firms in Pakistan, the study tests a moderated mediation model via partial least squares structural equation modelling (PLS-SEM) in WarpPLS 8.0.FindingsIt has been found that BDAEDC is a significant source of improved exploitative and exploratory learning, which are responsible for improved decision-making quality. Explorative learning ends up being a more influential mechanism. Bootstrapped mediation analysis confirms both learning orientations as key pathways. Top-management support of learning has a negative moderating effect on both the associations.Originality/ValueBy combining theory on dynamic capabilities with theory on organizational learning, digital capabilities for analytics integrate technological and organizational perspectives and create a new understanding of the mechanisms and boundary conditions of the effect of analytics-enabled capabilities on decision-making quality.
PurposeRecent studies have identified methodological concerns in data science research, including unvalidated preprocessing assumptions that may contribute to inflated performance claims and irreproducible results. Despite preprocessing consuming 60-80% of analytical effort in business analytics projects, strategy selection typically relies on convention rather than systematic evidence. This study introduces REPROPREP, a methodological framework designed to enable systematic validation of preprocessing effectiveness assumptions.Design/methodology/approachThe REPROPREP framework incorporates conservative statistical analysis with Benjamini-Hochberg false discovery rate correction, systematic quality degradation protocols, and cost-effectiveness assessment. The framework was evaluated across 10 UCI repository datasets using gradient boosting classifiers with 5-fold stratified cross-validation under three simulated quality conditions, resulting in 90 statistical comparisons.FindingsThe analysis found no statistically significant differences between preprocessing strategies after multiple comparisons correction, with negligible effect sizes (mean AUC difference: 0.001). Cost analysis indicated implementation cost differences ranging from $150 to $800 across strategies.Practical implicationsThe framework provides a systematic methodology for preprocessing evaluation, enabling organizations to conduct context-specific validation of preprocessing effectiveness assumptions and support more informed decision-making regarding preprocessing strategies.Originality/valueREPROPREP introduces a reproducible and systematic approach to preprocessing validation that integrates statistical rigor with cost-benefit analysis. The framework is designed for adaptive development, with this initial release focusing on numeric preprocessing and providing a foundation for future extensions.
PurposeThis study examines how emerging digital technologies-particularly artificial intelligence (AI), machine learning, big data analytics, and blockchain-are transforming tax audit practices. The research aims to synthesize existing evidence on the role of these technologies in improving audit efficiency, detection accuracy, and enforcement effectiveness across different jurisdictions.Design/Methodology/ApproachA systematic literature review was conducted using a PRISMA-guided methodology. Peer-reviewed publications published between 2017 and 2024 were collected and screened, resulting in a final sample of 23 studies. The selected studies were analyzed and synthesized to identify dominant technological applications, institutional contexts, and emerging research patterns in technology-enabled tax auditing.FindingsThe review identifies four major thematic clusters: AI-enabled automation in audit processes, blockchain-based audit integrity and traceability, data-analytics-driven risk detection, and regulatory and institutional readiness. The findings indicate that technologically advanced jurisdictions achieve significant improvements in audit precision, real-time anomaly detection, and transactional transparency. In contrast, developing economies face persistent challenges related to digital infrastructure limitations, technical capacity constraints, and fragmented regulatory frameworks.Originality/ValueThis study contributes to the tax administration literature by providing a comprehensive taxonomy of emerging tax audit technologies and offering a cross-jurisdictional synthesis of their effectiveness. It also proposes a conceptual framework integrating technological drivers, institutional moderators, and audit performance outcomes, thereby offering a structured perspective for analyzing digital audit transformation.Practical ImplicationsThe findings highlight important policy considerations for tax authorities and policymakers, particularly the need to strengthen data governance frameworks, invest in auditor digital competencies, and adapt regulatory systems to support the implementation of technology-enabled tax audit mechanisms.
PurposeAmid rising environmental expectations, green supply chains must balance profitability, sustainability, and fairness. This study examines how government subsidies, retailer green service investment, and fairness preferences jointly influence optimal pricing, product greenness, service effort, and profit allocation.Design/methodology/approachA Stackelberg game-theoretic model with a manufacturer and a retailer is developed. Four scenarios are analyzed: no intervention, subsidy only, service only, and combined subsidy-service. The model further incorporates fairness concerns of either party using an asymmetric Nash bargaining framework.FindingsThe results show that government subsidies improve product greenness and supply chain profitability, particularly when aligned with retailer-led green service investments. Service investments enhance demand but may reduce profits without policy support. Fairness preferences also reshape decision outcomes: manufacturer fairness raises prices but reduces efficiency, whereas retailer fairness promotes equity but may weaken upstream sustainability. A fairness sensitivity threshold enables profit coordination with minimal efficiency loss.Practical implicationsAligning public subsidies with retailer green service initiatives can improve both environmental performance and supply chain profitability.Originality/valueThis study integrates economic incentives and behavioral fairness considerations into a unified framework for analyzing sustainable supply chain strategies.
IntroductionMany retailers face growing pressure to enhance customer experience as consumer behavior shifts rapidly, especially after Covid-19. Shoppers now seek greater convenience and value, making it essential to understand not only what they buy and why, but the entire journey leading to their purchase decisions. AI provides powerful opportunities to transform retail from operational optimization to personalized engagement by improving and elevating the overall shopping experience.MethodologyHow can AI solutions be designed to help retailers understand in-store customer actions and identify different customer types for data-driven marketing? To address this question, this study proposes an AI-enabled track-and-trace framework that captures customer movements and behaviors through computer-vision AI, motion and emotion analysis, beacon technology, and various sensors.ResultsOur study focuses on the customer journey and then identifies different customer types based on their observed journey patterns. By monitoring customers in real time, the system supports personalized, location-based marketing strategies such as targeted follow-ups and customized advertising.Practical implicationsGiven the challenges of improving information visibility through these technologies, this study presents several use cases demonstrating how AI can incorporate the customer's perspective into marketing strategies to deliver more effective personalized offers and campaigns.
IntroductionIn complex operational environments, hybrid decision-making frameworks offer a means to integrate human expertise - characterized by contextual sensitivity, adaptability, and experiential knowledge - with the objective, standardized precision of machine-based systems.MethodThis study develops a decision-support structure by comparing supervised Machine Learning (ML) models; random forest (RF), support vector machine (SVM) and a bidirectional encoder representation from transformer-based (KB/BERT) model - using hierarchical and flat approaches against a manual classification process, involving more than 200 train delay codes across 10 days. ML models are trained on same-day delay data and evaluated against the outcomes from a multi-actor decision process.ResultsHierarchical models outperform flat ones, achieving near-human assessors on basic level coding (Level 1 and 2), though with greater variability (mean F1-scores (50-91 per cent)), compared to manual classification (mean F1-scores (87-98 per cent)) at the most granular level (Level 3) of prediction. "Simpler" models also outperform the more complex KB/BERT.Practical ImplicationsWe discuss the functionality and accuracy of ML-based hybrid decision-support systems (HDSS), noting the need for trade-offs between precision and accuracy. ML models demonstrate potential to complement - not replace - human expertise, particularly with uncertainty estimation tools that mitigate classification risks and support decision-making. We conclude with implications for data representation in the design of HDSS within socio-techno-economic contexts.
The rapid evolution of Artificial Intelligence (AI) has significantly influenced business decision-making across industries. This study conducts a systematic literature review to examine how elements of human consciousness are being integrated into AI frameworks to enhance decision-making quality and ethical awareness, with a specific focus on the Fast-Moving Consumer Goods (FMCG) sector. Using the PRISMA methodology, 40 peer-reviewed studies from domains including AI, cognitive science, and management were analyzed to synthesize theoretical and empirical insights. The review identifies four key thematic areas: (1) foundational theories and consciousness lenses in AI, (2) decision-making augmented by consciousness-inspired attributes, (3) cognitive architectures that emulate self-reflection and contextual reasoning, and (4) current AI adoption practices in FMCG. Findings reveal that, although true machine consciousness remains unrealized, integrating conscious-like attributes - such as introspection, situational awareness, and ethical deliberation - enhances the transparency and reliability of AI-assisted decisions. The paper concludes with a structured research agenda outlining future directions for developing consciousness-oriented AI systems for managerial decision-making.
Despite the immense importance of digital technologies, there has been a noticeable lack of attention to designing and implementing new evaluation models and frameworks for assessing their providers. To address this gap, an advanced Data Envelopment Analysis (DEA) model is developed to account for changes in both input and output variables. The developed model incorporates the Directional Distance Function (DDF) and non-radial efficiency functions, enabling more precise evaluations of Internet of Things (IoT) providers. Additionally, it classifies providers into three distinct categories: Pareto-efficient, weak-efficient, and inefficient, offering a clearer assessment of their overall efficiency. Overall, the model provides a unique and comprehensive framework for evaluating the sustainability and resilience of IoT providers. The sustainability and resilience of IoT providers are evaluated through the development of a novel analytical model. Based on the DDF, proportional inputs and outputs are considered within the DEA framework. A variety of data types, including integers and ratios, are integrated into hybrid returns to scale (HRS) technology. For the first time, directional distance and non-radial efficiency functions are incorporated into HRS technology. The proposed model can classify IoT providers into three categories: Pareto-efficient, weak-efficient, and inefficient.
Anomaly detection is an essential task for many firms and organizations. Identifying unusual patterns in messy multivariate time series can prevent catastrophic events and optimize operations. Traditional statistical methods struggle with high-dimensional data and complex temporal dependencies. In this paper, we propose a novel approach that combines Generative Adversarial Networks (GANs), a reconstruction-based framework, and Graph Neural Networks (GNNs) for effective and interpretable anomaly detection. Our method involves representing multivariate time series as graphs and training two interconnected GANs and an Autoencoder to capture the normal behaviour of the networks. Anomalies are detected by measuring the reconstruction error and the discriminator's score. Our case study on the signals of the INFN CNAF Tier-1 data centre demonstrates the effectiveness of our approach in terms of robustness and interpretability. We further validate our model on two widely used benchmark datasets for IT infrastructure monitoring, PSM and SMD, obtaining ${F_\beta }$F beta scores of 0.943 and 0.763, respectively. This work highlights the potential of GNNs in developing interpretable deep learning solutions for real-world applications.
IntroductionThis study introduces a framework for evaluating consistency in large language model (LLM) binary text classification, addressing the lack of established reliability assessment methods.MethodologyAdapting psychometric principles, we determine sample size requirements, develop metrics for invalid responses, and evaluate intra- and inter-rater reliability. Our case study examines financial news sentiment classification across 14 LLMs (including claude-3-7-sonnet, gpt-4o, deepseek-r1, gemma3, llama3.2, phi4, and command-r-plus), with five replicates per model on 1,350 articles.ResultsModels demonstrated high intra-rater consistency, achieving perfect agreement on 90-98% of examples, with minimal differences between expensive and economical models from the same families. When validated against StockNewsAPI labels, models achieved strong performance (accuracy 0.76-0.88), with smaller models like gemma3:1B, llama3.2:3B, and claude-3-5-haiku outperforming larger counterparts. All models performed at chance when predicting actual market movements, indicating task constraints rather than model limitations.Practical implicationsOur framework provides systematic guidance for LLM selection, sample size planning, and reliability assessment, enabling organisations to optimise resources for classification tasks.
This paper deals with the demand forecasting of consumer packaged goods. While Prophet and LSTM are established tools in time-series forecasting, known for their reliability and effectiveness over traditional techniques, the main goal achieved in this paper is the development of a hybrid Prophet-LSTM framework that outperforms both Prophet's and LSTM's capabilities, bridging the gap between time-series data mining and economic analysis. The methodology has been assessed on a large scale using a local modelling approach, outperforming Transformer-based architectures as well, with results confirmed by well-suited statistical tests. Furthermore, integrating the proposed framework with the PCHIP algorithm has led to a method to forecast demand curves of goods over a medium-term horizon. In addition, two reliability criteria have been assessed for the generated forecasts: an a posteriori qualitative criterion derived from the law of demand from microeconomics, and an a priori quantitative criterion based on the coefficient of variation metric. Accordingly, two model selection approaches have been introduced. Each involves switching to a more suitable deep learning model when the corresponding reliability criterion classifies Prophet's forecasts as not very reliable. Finally, the economic perspective has clearly emerged as a key factor guiding the choice between the proposed model selection strategies.
In this research, the authors map twenty years of luxury fashion marketing research (282 Scopus-indexed articles ranging from the year 2004 to 2023) using a novel combination of bibliometric analysis and BERT-based topic modelling. Eight major research themes encompassing thirty-seven granular topics are identified, offering a more detailed landscape than prior reviews. This transformer-driven BERTopic approach achieved substantially higher topic coherence than a classical LDA model, enhancing the reliability of insights. The authors further introduced an association rule mining-based "topic recommender" to reveal under-explored topic intersections. Findings highlight emerging focuses such as sustainability, digital luxury experiences and authenticity. It contributes to a comprehensive framework of luxury fashion marketing knowledge. The paper provides insights to managerial implications for luxury brand strategy related to leveraging social media engagement, safeguarding brand heritage, and embracing sustainable practices. It also proposes future research directions on AI-driven personalisation, the sustainability - exclusivity paradox and cross-cultural digital identities.
Background: With the global popularity of the Internet since the twenty-first century, the amount of data information has increased exponentially. Data mining and application have become one of the focuses of all circles. Data information can help users quickly find products that meet their interest needs through personalized recommendation, precision marketing, and other ways, so as to improve the shopping experience. It can also optimize the allocation of learning resources through data analysis, so that learners can acquire the required knowledge more effectively and improve learning efficiency. As a widely used algorithm in data utilization, recommendation algorithms can make accurate recommendations to users based on their past historical data. Previous recommendation algorithms, such as user-based collaborative filtering and item-based collaborative filtering, content-based recommendation technology, and matrix decomposition-based methods, have insufficient accuracy and slow computing speed, which cannot meet the needs of platforms and users. Method: Therefore, this paper proposes an online shopping platform product recommendation model to make the recommendation more accurate. This is a combination of the attention mechanism and GRU (Gated Recurrent Unit, GRU) algorithm. Result: Through the test on a real data set, the recognition rate of this research-proposed algorithm reaches 90.2% in 15 iterations, the average time of a single iteration is 21 s, and the accuracy range is 95.32 similar to 96.03%. Compared with Singular Value Decomposition, the Non-negative Matrix Factorization recommendation algorithm, and the Personal Rank algo-rithm based on random walk, the research method has better performance in the aspects of recommendation accuracy and recall rate. Conclusion: The attention-based GRU recommendation model effectively improves the accuracy and efficiency of recommendation systems. It outperforms traditional methods and is well-suited for large-scale, high-concurrency scenarios in online shopping platforms.
Text analysis is increasingly used across social science fields. We apply both supervised and unsupervised methods to open-text responses from a large employee survey, an underused but widely available source of organizational data. Using a design science approach informed by sensemaking theory, we examine how different analytical choices shape the extraction of employee concerns. Two findings stand out. First, employees use open-text fields to voice concrete, pragmatic issues that differ markedly from the more abstract questions asked in the closed-form questions. Second, an unsupervised Structural Topic Model performs especially well, producing coherent themes with minimal manual coding. Taken together, the results demonstrate how organizations can use open-text data to access overlooked forms of employee voice.
This paper addresses gaps in electricity price forecasting by providing a head-to-head comparison of eight forecasting approaches filtered on South Australian data: the Energy Supply Association of Australia (ESAA) and National Electricity Market (NEM) datasets. We evaluate six standalone models (XGBoost, SARIMA, LSTM, DeepAR, Prophet, and Random Forest) and two novel hybrid configurations: XGBoost - SARIMA and Random Forest - LSTM. A key contribution lies in leveraging large, high-resolution data to rigorously benchmark these algorithms based on forecasting accuracy, robustness to extreme events, and computational efficiency. The study highlights the significant impact of incorporating exogenous features, implicitly handled by advanced algorithms, on forecasting performance. Superior results from hybrid models, particularly the Hybrid RF-LSTM for wholesale price forecasting, underscore the value of combining feature-based ML with deep temporal learning to capture complex nonlinear relationships and temporal dependencies. For consumer price forecasting, leveraging insights from the more data-rich wholesale market with a hybrid approach proved more reliable, especially with limited secondary data. By integrating rigorous algorithmic benchmarking, tailored feature engineering, and practical recommendations, this research provides a valuable toolkit for navigating electricity market volatility in South Australia and offers a replicable framework for other subregional markets.
Stock prediction is essential for informed investment decisions, risk management, and economic planning. This study aims to forecast stock price movements across the top eight Indian sectors using a combination of four multidimensional input types - numerical data, technical patterns, technical indicators, and textual data - within a unified predictive framework. The analysis focuses on eight major Indian sectors: Auto, Bank, Financial Services, FMCG, Metal, IT, PSU Bank, and Realty, over the period from 2017 to 2021. The study employs deep learning models including LSTM, GRU, CNN, Bi-LSTM, and a hybrid LSTM-GRU architecture on the integrated dataset. Model performance is evaluated using RMSE, MdSE, MAE, and R2. Among the models tested, the hybrid LSTM-GRU model achieved the highest prediction accuracy, with a peak performance of 98% in the IT sector. Studies that combine all four input types within a single predictive framework are rare, and the integration of an extensive range of technical patterns along with sector-wise analysis is seldom addressed in the existing literature. This study offers a more comprehensive and multidimensional approach to stock prediction, contributing valuable insights for investors, researchers, and policymakers.
Data storytelling combines narrative techniques with data visualisation to present complex information in an engaging and actionable manner. Recently, the potential use of Generative Artificial Intelligence (AI) in data visualisation and storytelling has gained attention but has seen limited empirical evaluation. This study addresses this gap by investigating whether Large Language Models (LLMs) can effectively assist individuals in crafting data-driven stories. To assess this, we designed a three-stage survey that captures participants' initial perceptions of Generative AI tools, followed by a hands-on activity where participants create data stories using LLMs, and concludes with a post-activity survey to evaluate changes in their attitudes. Our findings indicate that individuals with prior experience using Generative AI are more inclined to adopt these tools for data storytelling. However, those who have specifically used LLMs for visualisation purposes exhibit less enthusiasm towards adoption, potentially due to prior challenges or limitations encountered. Additionally, participants found Generative AI tools particularly useful in the initial stages of data storytelling, especially when identifying the target message. Importantly, we observed a significant positive shift in participants' likelihood to adopt Generative AI following hands-on experience. These insights highlight the potential of Generative AI in enhancing data storytelling processes and highlight opportunities for improved human-AI collaboration, particularly within Business Analytics.
This study investigates the relationship between the forecasting power of web-based financial news sentiments and market capitalisation by analysing large, mid, and small-cap Indian stocks at both the index and sectoral levels. The analysis is further extended to two emerging Asian economies, Malaysia, and Vietnam. Given the dominant contribution of the finance sector, the study analyses the impact of financial news sentiments within the financial sector across each market cap. For this, 1,54,448 news headlines are extracted from an online news website, where financial news is identified using a TFIDF-GRU model and a keyword search with 187 financial terms. Sentiments are computed using a hybrid Doc2Vec-TFIDF feature extraction technique and an SVM classifier. The study employs a hybrid BiLSTM-GRU model incorporating web-based financial news sentiments alongside technical and macroeconomic indicators such as 10-year bond yield, exchange rate, gold price, crude oil price, and S&P500 closing price. Findings reveal that the forecasting power of web-based financial news sentiments varies significantly with market cap, with a strong impact on large-cap and mid-cap stocks. The study holds significant economic and policy implications, offering actionable insights for stakeholders across financial markets, regulatory bodies, and government.
The rise of online social networks has created a new landscape for brands and companies to engage with consumers. Influencer marketing has emerged as a modern strategy that utilizes social media influencers to connect with target audiences quickly and effectively. The challenge for companies and decision-makers is how to determine and profile social media influencers, by extracting and classifying the opinions and topics they discuss. A limited number of studies suggest advanced methods for identifying social media influencers. The aim of this research is to introduce a new approach, Identifying and Profiling Influencers in the context of social media Marketing (IPIM) to determine influencers based on content mining on Twitter. The IPIM approach consists of four phases: (1) employing topic modeling to uncover relevant topics, (2), assembling topics into domains and marketing strategies, (3) categorizing the expertise levels using clustering techniques and (4) identifying and profiling marketing influencers with the PageRank algorithm and social network data. The findings show that an influencer can engage in multiple marketing domains and strategies. Moreover, the impact of influencers does not only rely on their fame or the size of their network. The IPIM approach demonstrates that a focused and relevant content is a key influence factor in the marketing field.
