East Delta University

This paper applies Machine Learning (ML) and explainable AI (XAI) to predict Environmental Social and Governance (ESG) controversies by using 9,609 firm-year observations from 3,194 leading companies from 27 countries in five regions, over the period of 2002 to 2023. We run Random Forest (RF), eXtreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), Support Vector Regression (SVR), Lasso and Ridge regression on high dimensional dataset with firm level ESG scores, financial and governance indicators, country level governance and macroeconomic indicators, and global ESG uncertainty indicator. Our results indicate that XGBoost and LightGBM outperform other Machine Learning (ML) algorithms and linear regression in training and testing data respectively. The predictive power and accuracy remain consistent across regions, countries and sub-sample period. Application of XAI confirms that ESG combined score is the most important feature to predict ESG controversies. Moreover, ESG sub-pillar scores are also important features to predict ESG controversies in global, regional and country setting. Further analysis shows that firms with higher ESG combined scores tend to have higher ESG controversies score i.e., lower ESG controversies. However, increase in individual sub-pillar scores lead to increase ESG controversies. Importantly, higher country level governance score, firms’ financial indicators and gender diversity can also predict ESG controversies accurately and reduce it to a certain extent. We recommend a number of policy measures for investors, regulators and policy makers.

The widespread use of online social media platforms has amplified the importance of efficient hate speech detection, especially in low-resource languages like Bengali. While traditional machine learning approaches show promise, deep learning is more effective in capturing the nuanced context of hate speech. Current challenges include a lack of diverse datasets and models capable of context-sensitive detection. To address these, we introduce HateCorpBN-XL, the largest labeled Bengali hate speech dataset to date, containing 65,251 comments across five categories: political (PoHS), religious (ReHS), misogynistic (MisoHS), slander (SlaHS), and xenophobic (XenHS). We also propose HateBertBN, a hybrid transformer-based model combining BanglaBERT embeddings with three neural network fusion strategies using CNN, LSTM, and MLP. We evaluate our approach on two tasks, Task-1: detecting hate speech in Bengali text classifying it as hateful or non-hateful and Task-2: categorizing hateful content into five distinct classes. For Task-1, all HateBertBN variants outperformed current transformer models, achieving an accuracy of 0.92 and a weighted F1-score of 0.92. In Task-2, the HateBertBN-MLP and HateBertBN-CNN variants achieved a notable 0.90 accuracy and weighted F1-score of 0.90, surpassing M-BERT, Distil-M-BERT, BanglaBERT, and XLM-R-Base. Although HateBertBN-LSTM performed slightly lower overall, it achieved strong F1-scores in the ReHS (0.93) and XenHS (1.00) categories. Overall, our hybrid model outperforms state-of-the-art approaches in both tasks, demonstrating its effectiveness and robustness.

PurposeThis study aims to examine the impact of digital accounting system quality on decision-making quality within the banking industry in Bangladesh. Specifically, it investigates how data accuracy and system efficiency influence decision quality by mediating information clarity and the moderating role of analytical decision orientation.Design/methodology/approachUsing a quantitative approach, data were collected from 287 respondents across various banking institutions in Bangladesh. Structural equation modelling was used to analyze the proposed mediated-moderated model. Measurement validity and reliability were confirmed through confirmatory factor analysis, and hypotheses were tested via bootstrapping techniques.FindingsThe results show that data accuracy significantly improves decision quality directly and indirectly through information quality. Although system efficiency does not directly affect decision quality, it significantly enhances information clarity, which in turn contributes to better decision outcomes. Furthermore, analytical decision orientation positively moderates the relationship between decision accuracy and information clarity, highlighting the role of organizational culture in decision-making effectiveness.Practical implicationsThis research holds substantial international relevance by offering a framework that financial institutions worldwide can adopt to assess and improve the effectiveness of digital accounting systems. The findings highlight the global importance of prioritizing data and information quality and fostering an analytical decision-making culture to ensure efficient decision outcomes in the digital era.Social implicationsBy enhancing transparency and decision accountability, effective digital accounting systems can contribute to improved financial governance and institutional trust. In the broader context of developing economies, where banking systems are evolving rapidly, empowering decision-makers with reliable, data-driven insights can help reduce errors, fraud and inefficiencies. Furthermore, cultivating a culture of analytical orientation supports ethical decision-making and strengthens organizational resilience, which has positive ripple effects on economic stability and public confidence in financial institutions.Originality/valueThis study contributes to the accounting literature by providing a context-specific understanding of how digital accounting systems influence decision-making in a developing economy. It fills a theoretical gap by modelling the combined mediating and moderating effects underexplored in prior studies.

Parkinson’s disease (PD) poses a growing global health challenge, with Bangladesh experiencing a notable rise in PD-related mortality. Early detection of PD remains particularly challenging in resource-constrained settings, where voice-based analysis has emerged as a promising non-invasive and cost-effective alternative. However, existing studies predominantly focus on English or other major languages; notably, no voice dataset for PD exists for Bengali – a language spoken by over 230 million people worldwide – posing a significant barrier to culturally inclusive and accessible healthcare solutions. Moreover, most prior studies employed only a narrow set of acoustic features, with limited or no hyperparameter tuning and feature selection, and little attention to model explainability. This restricts the development of a robust and generalizable machine learning (ML) model. To address this gap, we present BenSparX, the first Bengali conversational speech dataset for PD detection, along with a robust and explainable ML framework tailored for early diagnosis. The proposed framework incorporates diverse acoustic feature categories, systematic feature selection methods, and state-of-the-art ML classifiers with extensive hyperparameter optimization. Furthermore, to enhance interpretability and trust in model predictions, the framework incorporates SHAP (SHapley Additive exPlanations) analysis to quantify the contribution of individual acoustic features toward PD detection. Our framework achieves state-of-the-art performance, yielding an accuracy of 95.67%, F1 score of 95.62%, and AUC of 0.990. We further validated our approach by applying the framework to existing PD datasets in other languages, where it consistently outperforms state-of-the-art approaches. This study lays the foundation for identifying subtle yet clinically meaningful vocal biomarkers, particularly in low-resource settings such as Bengali-speaking populations, and represents a significant step toward equitable, explainable, and robust digital health diagnostics for neurodegenerative disorders.

The study investigates the association between hotel marketing mix, tourist satisfaction, revisit intention in the Bangladesh hotel industry, highlighting the mediating and moderating effects of satisfaction and social influence. Data were collected from 404 tourists who stayed in upscale hotels (3, 4, and 5-star) in Chittagong and Cox's Bazar. The gathered data were assessed and interpreted using Smart PLS software. Tourist satisfaction significantly mediates the relationship between the hotel product, price, location, promotion and revisit intention. In addition, social influence moderates tourists' satisfaction and revisits the intention relationship. The framework and findings can serve as a strategic tool to expedite hotel top management and managers' decision-making, aiming to improve tourist revisit intention in the competitive business environment through hotel marketing mix and satisfaction. The findings contribute to the Theory of Planned Behaviour (TPB) by explaining the mediating role of satisfaction and the moderating role of social influence in augmenting the intention to revisit in the Bangladeshi hospitality industry.