This study challenges the prevailing belief in the necessity of complex models for accurate forecasting by demonstrating the effectiveness of parsimonious econometric models, namely ARCH(1) and GARCH(1,1), over deep learning robust approaches, such as LSTM and 1D-CNN neural networks, in modeling historical volatility within pre-emerging stock markets, specifically the Moroccan and Bahraini stock markets. The findings suggest reevaluating the balance between model complexity and predictive accuracy. Future research directions include investigating the potential existence of threshold effects in market capitalization for optimal model performance. This research contributes to a deeper understanding of volatility dynamics and enhances forecasting models’ effectiveness in diverse market conditions.

Recent studies underscore the growing application of machine learning (ML) in finance, as revealed through bibliometric analyses. Kureljusic and Karger (2024) reviewed AI-based forecasting in financial accounting, identifying gaps and proposing future research agendas. Similarly, Biju et al. (2023) explored the taxonomy of AI, deep learning, and ML in finance, highlighting the rise in publications and the need for empirical studies on algorithmic financial technologies. Building on this foundation, this study presents a scientometric analysis of ML in financial data forecasting from 1996 to 2024. Using data from Scopus and Web of Science, we examine key themes, collaboration networks, and influential contributors. Employing deep learning tools like Sentence-BERT, BerTopic, BERT, ChatGPT, and PEGASUS, we offer insights into how ML has reshaped financial forecasting. This analysis provides a basis for future research, guiding scholars and practitioners towards impactful areas in finance.

The proliferation of AI-generated content has significantly impacted the advertising and entertainment industries. This study investigates the influence of AI-generated influencer content on consumer engagement. Employing a qualitative research methodology, we analyze a large dataset of social media posts featuring both human and AI-generated influencer content. Through deep learning-based text analysis and sentiment analysis, we aim to identify patterns, themes, and consumer responses to these different content types. The findings will contribute to a deeper understanding of the evolving landscape of influencer marketing and provide insights for brands and marketers.

Forecasting association football is hard for reasons that are structural rather than incidental: scoring is sparse, international fixtures are infrequent and non-stationary, and many tournament pairings share little competitive history. Because the outcome of a match is irreducibly random, the appropriate object of prediction is not a single result but a probability distribution over results. This paper develops a compact, fully interpretable model that targets that distribution directly. Team strength is summarised by a rolling Elo rating; the rating gap between two sides is mapped through a log-linear link onto a pair of Poisson goal rates; and the joint scoreline distribution is obtained from the bivariate adjustment of Dixon and Coles (1997), which corrects the systematic under-prediction of low-scoring draws under independent Poisson sampling. The four scalar parameters of the scoring layer are estimated from 28 644 historical internationals rather than assumed, and the fitted model is benchmarked out-of-sample against a flexible gradient-boosting classifier. The parametric model performs comparably to the benchmark while remaining considerably more interpretable, suggesting that additional model flexibility does not materially improve out-of-sample performance for this information set. Match-level distributions are propagated through the competition by Monte-Carlo simulation to obtain advancement and title probabilities. Applied to a supplied Round-of-32 bracket, the model identifies Spain (≈29%) and Argentina (≈ 27%) as the leading contenders and isolates the fixtures in which an upset is most plausible.

Quantitative portfolio optimization has accelerated sharply since 2018, with deep learning and reinforcement learning agents now competing with the mean-variance framework that defined six decades of research. Existing narrative reviews struggle to track this expansion. We screen 832 documents from Scopus and Web of Science under PRISMA 2020 and retain 589 unique articles spanning 2003–2025. Applying BERTopic with SPECTER scientific embeddings, UMAP and HDBSCAN, we identify five coherent topics with mean coherence 0.864: classical mean-variance (T0; n = 270), deep reinforcement learning (T1; n = 116), machine-learning return forecasting (T2; n = 87), covariance estimation and robust optimization (T3; n = 52) and metaheuristics (T4; n = 56). A rank-weighted similarity analysis, designed to neutralise the c-TF-IDF collinearity artefact, shows that deep reinforcement learning is the most isolated paradigm. The two methodological families bifurcate over time: AI / deep-learning approaches grow from 3.6 % of annual output before 2018 to 40.2 % afterwards, while classical methods retain volume but lose share. We synthesise the empirical practices of each family along five dimensions critical to applied finance and identify three under-explored integration frontiers.

Quantitative portfolio optimization has accelerated sharply since 2018, with deep learning and reinforcement learning agents now competing with the mean–variance framework that defined six decades of research. Existing narrative reviews struggle to track this expansion. We screen 832 documents from Scopus and Web of Science under PRISMA 2020 and retain 589 unique articles spanning 2003–2025. Applying BERTopic with SPECTER scientific embeddings, UMAP and HDBSCAN, we identify five coherent topics with a mean coherence of 0.864: classical mean–variance (T0; n = 270), deep reinforcement learning (T1; n = 116), machine learning return forecasting (T2; n = 87), covariance estimation and robust optimization (T3; n = 52)—and metaheuristics (T4; n = 56). A rank-weighted similarity analysis, designed to neutralise the c-TF-IDF collinearity artefact, shows that deep reinforcement learning is the most isolated paradigm. The two methodological families bifurcate over time: AI/deep learning approaches grow from 3.6% of annual output before 2018 to 40.2% afterwards, while classical methods retain volume but lose share. We synthesise the empirical practices of each family along five dimensions critical to applied finance and identify three under-explored integration frontiers.