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.

Generative AI coding assistants are increasingly used to write machine-learning code, yet their ability to produce reliable LSTM implementations for financial prediction remains underexplored. This study evaluates the LSTM code generated by seven assistants ChatGPT 4.5, GitHub Copilot, Deepseek 3, Perplexity, Gemini 2.0 Pro, Claude 3.7 Sonnet, and Meta’s Llama from a single standardized prompt, on three indices (Nikkei 225, S&P 500, STOXX Europe 600). Each assistant’s generated script was re-executed over independent runs; accuracy (MAE, MSE, RMSE, R ² , execution time) is reported as mean ± standard deviation on the original price scale, complemented by a static code-quality analysis (Pylint, Radon, SonarQube, Pytest, Bandit). The assistants converge on nearly identical LSTM architectures, so performance differences arise mainly from data-handling and code-correctness defects: Meta’s Llama near-zero errors are an artifact of normalized-scale metrics combined with a shuffled train/test split (data leakage), and once corrected its accuracy is among the weakest; Gemini 2.0 Pro, once its predictions are evaluated consistently on the price scale, is among the most accurate assistants. Differences are validated with Diebold–Mariano and Wilcoxon tests. AI-generated forecasting code can be accurate but is not uniformly trustworthy: its generated preprocessing and evaluation code must be audited before use.

Emerging markets, particularly in Africa, present unique challenges for portfolio optimization due to increased volatility and limited historical data. This study explores the potential of deep learning models, specifically Long Short-Term Memory (LSTM), Deep Multilayer Perceptron (DMLP), and Convolutional Neural Networks (CNN), to predict the movements of three key African stock indices: MASI (Morocco), BRVM Composite (West Africa) and TUNINDEX (Tunisia). By comparing the performance of these architectures, this research aims to identify the models most suitable for prediction in a the context of limited data and high volatility. The results of this study will provide crucial insights for investors, portfolio managers, and researchers, thereby contributing to the development of more robust and effective portfolio optimization strategies for African markets, often overlooked by traditional methods.

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.

The auditing of predictive models in finance requires balancing rigorous quantitative assessment with robust algorithmic governance. This chapter examines the evaluation of advanced machine learning models—LSTM, DMLP, CNN, and XGBoost—applied to financial markets, including the FTSE CSE Morocco 15 (2018–2023), the S&P 500, and the STOXX Europe 600. Models were assessed using both quantitative metrics (R2, RMSE, MAE) and qualitative controls to ensure transparency, reliability, and regulatory compliance, which are critical for investor and regulator trust. Empirical results indicate that XGBoost achieved the highest stability and accuracy (R2 up to 0.91; RMSE standard deviation < 3%), followed by LSTM (R2 up to 0.88; standard deviation ≤ 4%), while CNN and DMLP exhibited higher variability (6%–7%) and overfitting. Bias audits and explainability analyses using SHAP and LIME confirmed the absence of systematic deviations and highlighted trade-offs between interpretability and performance. A criticality matrix based on ISO 31000/23894 standards guided the prioritization of mitigation plans, supporting operational deployment with continuous monitoring, drift detection, and periodic recalibration. This study provides a comprehensive framework for auditing predictive models in finance, integrating technical evaluation with governance principles to enhance transparency, trust, and accountability in algorithmic decision-making.