The paper analyzes the volatility trends of the Saudi Arabian Tadawul All Share Index (TASI) and the S&P 500 index, focusing on the COVID-19 pandemic as a key market shock. The analysis incorporates daily stock return data covering the period from January 2015 to May 2025. The volatility of emerging and developed markets is examined through EGARCH and GARCH approaches to study characteristics such as volatility clustering and asymmetry. The effect of the pandemic is directly embedded by introducing COVID-19 dummy variables into the models. Empirical findings suggest that both indices are characterized by volatility clustering, and the EGARCH model is more appropriate than the GARCH model for estimating asymmetric volatility, particularly during crisis periods. Additionally, the COVID-19 dummy variable is statistically significant in the EGARCH model, as opposed to the GARCH model. The results support the leverage effect, indicating that negative shocks have a more significant impact on market volatility than positive ones. The S&P 500 showed a faster recovery after the COVID-19 crisis, whereas TASI was slower in mean reversion, indicating structural and behavioral divergence between the markets. This comparative study contributes to the literature by providing a clear picture of volatility dynamics in diverse financial contexts and highlighting the superiority of EGARCH models during crisis periods. The findings offer guidance to policymakers aiming to improve market stability and to investors seeking diversification into both developing and mature markets.

Abstract: Machine learning is transforming stock market prediction by leveraging vast datasets, advanced algorithms, and computational power to provide more accurate forecasts. While challenges remain, continuous advancements in artificial intelligence and deep learning are improving predictive models, making them an essential tool for traders and investors. Stock market prediction extremely challenging due to the dependence of stock prices on several financial, socio-economic and political parameters etc. For real life applications utilizing stock market data, it is necessary to predict stock market data with low errors and high accuracy. This needs design of appropriate artificial intelligence (AI) and machine learning (ML) based techniques which can analyze large and complex data sets pertaining to stock markets and forecast future prices and trends in stock prices with relatively high accuracy. This paper presents a comprehensive review on the various techniques used in recent contemporary papers for stock market forecasting. Keywords: Time Series Models, Stock Market Forecasting, Artificial Intelligence, Artificial Neural Networks, Forecasting accuracy.

ABSTRACT This paper constructs a new investor attention index utilizing search information for S&P 500 index-linked exchange-traded funds. The proposed index demonstrates strong positive predictive power for aggregate stock market excess returns over a forecasting period of up to two years, both in- and out-of-sample. This index has significant superiority over previously examined macroeconomic variables, and its predictive power is comparable and is informationally complementary to prevailing attention and sentiment indices. In addition, the predictive models relying on the newly proposed index can yield sizable economic gains for a mean-variance investor. Finally, we find that the new investor attention index’s forecasting ability predominantly stems from its capacity to explain future cash flow news and high-variance stock returns.

Financial markets generate complex, high-dimensional time series characterized by nonlinearity, noise, and structural uncertainty, posing persistent challenges for classical machine learning and statistical forecasting models. This paper examines quantum machine learning (QML) architectures for stock market forecasting, anomaly detection, and pattern recognition in financial time series. By leveraging quantum phenomena such as superposition, entanglement, and quantum-enhanced feature spaces, QML models offer new computational paradigms for capturing intricate market dynamics and latent structures. The study synthesizes recent advances in quantum neural networks, variational quantum circuits, and hybrid quantum–classical models, evaluating their applicability to predictive modeling and risk-sensitive detection tasks in finance. The paper argues that QML architectures have the potential to complement classical approaches by improving representational capacity and scalability for complex financial data, while also outlining current technical limitations and research directions necessary for practical deployment.

Financial markets generate complex, high-dimensional time series characterized by nonlinearity, noise, and structural uncertainty, posing persistent challenges for classical machine learning and statistical forecasting models. This paper examines quantum machine learning (QML) architectures for stock market forecasting, anomaly detection, and pattern recognition in financial time series. By leveraging quantum phenomena such as superposition, entanglement, and quantum-enhanced feature spaces, QML models offer new computational paradigms for capturing intricate market dynamics and latent structures. The study synthesizes recent advances in quantum neural networks, variational quantum circuits, and hybrid quantum–classical models, evaluating their applicability to predictive modeling and risk-sensitive detection tasks in finance. The paper argues that QML architectures have the potential to complement classical approaches by improving representational capacity and scalability for complex financial data, while also outlining current technical limitations and research directions necessary for practical deployment.

Abstract This research proposes an innovative Arabic financial forecasting model that integrates linguistic relation extraction with advanced deep learning and machine learning techniques. The framework combines Arabic Open Information Extraction (AOIE), BERT-based contextual sentiment analysis combined with zero-shot XGBoost, forming an end-to-end architecture capable of interpreting both the semantics and emotions of Arabic financial text. The model aims to address the linguistic and resource challenges inherent in Arabic financial data by leveraging syntactic and semantic structures extracted from unstructured sources such as news reports and financial statements. Through extensive experiments, the proposed approach demonstrated consistent and superior predictive performance across different configurations. The optimal setting achieved an accuracy of 97.4%, with a Mean Absolute Error (MAE) of 0.13 and a Root Mean Square Error (RMSE) of 0.18, confirming its reliability and robustness in forecasting stock market trends. Compared to traditional statistical models (ARIMA, VAR) and deep learning baselines (LSTM, CNN, Transformer-only), the proposed AOIE–BERT–zero-shot XGBoost framework achieved the lowest prediction error and highest interpretability. The findings underscore the significance of incorporating Arabic linguistic structures into predictive modeling and demonstrate the potential of transformer-based NLP integration for financial analytics. This research contributes a scalable and linguistically adaptive solution, paving the way for more accurate, explainable, and multilingual applications of Natural Language Processing (NLP) in the global financial domain.

Forecasting stock market crashes is difficult because such events are rare, highly nonlinear, and shaped by latent structural and behavioral forces. This study introduces a calibrated and interpretable Random Forest framework for detecting pre-crash conditions through structural feature engineering, early-warning calibration, and model explainability. Using daily data on global equity indices and major large-cap stocks from the U.S., Europe, and Asia, we construct a feature set that captures volatility expansion, moving-average deterioration, Bollinger Band width, and short-horizon return dynamics. Probability-threshold optimization significantly improves sensitivity to rare events and yields an operating point at a crash-probability threshold of 0.33. Compared with econometric and machine learning benchmarks, the calibrated model attains higher precision while maintaining competitive F1 and MCC scores, and it delivers meaningful early-warning signals with an average lead-time of around 60 days. SHAP analysis indicates that predictions are anchored in theoretically consistent indicators, particularly volatility clustering and weakening trends, while robustness checks show resilience to noise, structural perturbations, and simulated flash crashes. Taken together, these results provide a transparent and reproducible blueprint for building operational early-warning systems in financial markets.

The Role of Internet of Things Data in Stock Market Forecasting: A New Frontier for Financial Management

With the mixing of machine learning (ML) algorithms, stock market forecasting has become an area of interest for researchers. This paper presents a review of the literature on ML techniques for price prediction in Indian stock markets. It analyses major algorithms such as linear regression, support vector machines (SVM), random forest, and neural network, as well as discussing their theoretical, advantageous and disadvantageous aspects. This research offers an exhaustive assessment of the implementation of diverse ML techniques for predicting trends in the Indian stock market. The swift changes and the nonlinear nature of stock markets make the problem of prediction extremely difficult. The author demonstrates how ML techniques are used to improve the accuracy of predictions by analysing historical data, sentiments of the market, and various technical indicators. Furthermore, it discusses issues such as crash in data, unpredictability of the market, and overfitting that create challenges for forecasting in ML. This review is aimed at providing substantial evidence in the implementation of various ML techniques adjusted to the Indian stock market for predicting stock price movements and providing information where further research is needed.

Wavelet-enhanced multimodel framework for stock market forecasting: A comprehensive analysis across market regimes

Since the inception of stock trading, scholars and investors have searched for reliable methods to forecast the course of stock values the next day. Since there are several variables that might influence the stock values of the next day, forecasting stock prices is a challenging undertaking. Stock Market Forecasting (SMF) is a forward-looking process anticipating future stock values, allowing to make sound financial decisions. In order to create predictions, academics and investors have started using machine learning approaches in conjunction with technical indicator analysis. However, the precision of the predictions is lacking. One of the progress in applying ML, particularly LSTM networks, to stock market forecasting lies in automating this process. Human bias implies that the same predictions can be misleading and contribute to the fact that they need to use ML and AI technology. The data used was fetched from finance.yahoo.com, and for confidence in the data, it took steps such as lemmatisation, null value management and deletion of duplicates. A total of four different ML prediction methods were utilised: LSTM is also being used ANN, CNN, K-Nearest Neighbour and many other algorithms. The model's performance was evaluated using measures including F1-score(Fs), recall(Rc), accuracy(Acc), and precision(Pr). Outcomes showed that the models were not all equally successful; however, the LSTM model had the best accuracy at 93%. Future attempts might consider other categorisation strategies and improving preprocessing methods to improve model performance and forecast Acc

Forecasting stock prices remains a highly challenging problem due to the volatile, nonlinear, and unpredictable nature of financial markets. Traditional forecasting models often fail to effectively integrate real-time data, capture nuanced sentiment, or adapt to rapidly changing market dynamics. To address these limitations, this paper proposes the Intelli Fusion Adaptive Decision Engine (IADE), a comprehensive hybrid framework that unifies multiple advanced AI technologies, including Deep QLearning (DQN), the Prophet time-series algorithm, Bidirectional Encoder Representations from Transformers (BERT), Adaptive Resonance Theory Neural Networks (ART-NN), and Transformer-based architectures with attention mechanisms. IADE is designed to improve user accessibility, enhance real-time forecasting accuracy, increase sentiment analysis precision, and enable adaptive predictive behavior. Experimental results demonstrate that the proposed system substantially improves forecasting performance and strengthens decision-making effectiveness in highly volatile financial environments.

Forecasting stock market trends is notoriously difficult because financial markets are intrinsically volatile and do not follow linear patterns. This project aims to address the problem of predicting the future direction of Netflix (NFLX) stock price, a representative stock with a significant position in the streaming industry. This study utilized daily historical stock price data from May 2002 to October 2024 and constructed a feature set comprising 20 technical indicators. The primary objective of this research is to construct a binary classifier that determines if the stock value will rise over 1% in a t+5 timeframe. To achieve this, this study compared four different machine learning and deep learning models: Logistic Regression, Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Long Short-Term Memory (LSTM). The experimental results show that all models provide predictive capabilities superior to random guessing. Among them, the LSTM model performed the best, achieving an accuracy of 53.73% on the test set, closely followed by XGBoost (53.05%). This result shows the possibility of deep learning models in capturing complex time-series dependencies.

Stock market forecasting represents a cornerstone of modern finance. In this field, the ability to predict even marginal improvements in forecasting accuracy translates to billions in enhanced returns and reduced losses. Recently, the emergence of machine learning techniques has challenged the Efficient Market Hypothesis (EMH). This hypothesis predicts that the developed markets should demonstrate a lower predictability than emerging markets due to superior information efficiency. While research related to market forecasting using novel models has proliferated, the existing literature lacks exploration of its effectiveness in comparison with other statistical methods and across different economic entities. To address these gaps, this paper presents an empirical framework that systematically evaluates forecasting performance across eight major economies representing both developed and emerging markets using a rolling window cross-validation approach. Using World Bank Global Economic Monitor data from the 21st century, this research tested seven distinct models, ranging from Random Walk benchmarks to traditional econometric approaches and advanced machine learning techniques. As a result, the implementation reveals a paradoxical finding: emerging markets demonstrate systematically higher forecasting errors compared to developed markets, contradicting theoretical predictions that less efficient markets should be more predictable.

This study addresses the critical challenge of forecasting stock market movements by leveraging sentiment analysis of financial news headlines. Predicting the Dow Jones Industrial Average (DJIA) is of great significance to investors and financial institutions, as market trends are often influenced by public sentiment and rapidly evolving news cycles. Traditional quantitative models often struggle to capture the nuanced impact of textual information, especially in the context of imbalanced data distributions where rare but impactful market events are underrepresented. In this work, the paper constructs a comprehensive sentiment analysis framework utilizing advanced natural language processing methods to classify the sentiment of publicly available news headlines and examine their relationship with subsequent DJIA price fluctuations. By integrating headline sentiment scores with historical price data, our experiments systematically evaluate the predictive reliability of various machine learning models, with a particular focus on class imbalance mitigation. The results demonstrate that models employing class weighting in LightGBM outperform those using conventional resampling techniques, achieving recall rates of 0.45 for downturn prediction versus 0.12 with baseline methods. These findings highlight the value of algorithmic enhancements for rare event forecasting and suggest that richer, domain-specific representations may further improve predictive accuracy. Future research will explore enhanced features and time-series modeling to boost the robustness of financial sentiment analysis.

Stock market prediction has long been a challenging problem in the field of finance and investment. Accurately predicting the movements of stock prices is crucial for making informed decisions and maximizing investment returns. Traditional models mainly use historical prices. We found that there is a gap in research in integrating financial news into the model, which has emerged as a promising direction in enhancing predictive accuracy. This research aims to address this problem by exploring a multimodal approach by combining companies’ news articles and their historical stock data to predict future stock movements. The objective was to compare the performance of a Graph Neural Network (GNN) model with an LSTM model. The methodology employed in this research involves an LSTM model that embeds the historical data for each company and a language model to embed news articles. These embeddings will represent nodes that have relationships presented by edges within a graph. Using a GNN message aggregation technique known as GraphSAGE, the model should be able to capture interactions and dependencies between news articles, companies, and industries and use this information to predict future stock movements. Two target variable approaches are explored: one focusing on the binary classification of whether the stock price will increase or decrease, and the other considering the significance of the increase. This methodology was evaluated on two datasets, the US equities dataset and the Bloomberg dataset. The results showed that the GNN model was able to achieve better performance than the baseline LSTM model on both datasets. The GNN model achieved an accuracy of 53% on the first target, a statistically significant 1% improvement over the baseline, and a 4% precision gain on the second target, which confirms the effectiveness of exploiting financial news using graph-based models. Furthermore, we observed that increasing the number of news samples led to improved accuracy. We also find that headlines contain stronger predictive signal than full articles which is consistent with evidence that headlines disproportionately shape readers’ judgments and market reactions.

Financial forecasting that is both accurate and comprehensible, is crucial for stock market risk management. Many of the innovative approaches that have surfaced in recent years make use of deep learning, ensemble methods, and ensemble techniques that combine many approaches on financial data. Although these techniques frequently produce impressive results, they frequently function as "black boxes," making it challenging for people to examine how predictions are made because of a lack of transparency that breeds mistrust. This review examines studies that combine a variety of models. It makes things more readable and trustworthy by utilizing XAI, deep learning, and machine learning. LSTM, BiLSTM, CNN, XGBoost, ARIMA, and Prophet are a few of these. These are used to identify important trends in financial data and to monitor patterns over time. In addition to SHAP, LIME, and Layer-wise Relevance Propagation, there are more XAI technologies that can be useful. It displays the factors that assist models in making predictions. These justifications encourage more individuals to have faith in the outcome. The review discusses major issues, such as making models easy to comprehend, how well these models hold up in new marketplaces, and leveraging people's emotions when forecasting.

Forecasting stock market anomalies in emerging markets: An OPTUNA-optimized isolation forest and K-means approach

Hybrid CNN–LSTM Integrated with Temporal Fusion Transformer for Accurate and Interpretable Stock Market Forecasting

The stock market, a cornerstone of the global financial system, is characterized by its dynamic and volatile nature, which makes accurate price-trend prediction challenging. However, traditional statistical models often fail to capture this complexity. Recent advancements in Artificial Intelligence (AI), particularly Machine Learning (ML) and Deep Learning (DL), have transformed stock market forecasting by using diverse datasets and algorithms. This review examines recent studies on AI methodologies for stock market price trend prediction models by analyzing architectures, datasets, performance metrics, and limitations, with a focus on hybrid models, sentiment analysis, and dataset diversity. Hybrid approaches, including the Multi-Model Generative Adversarial Network Hybrid Prediction Algorithm (MMGAN-HPA), K-means long short-term memory (LSTM), and LSTM autoregressive output (LSTM-ARO), improve predictive accuracy by combining statistical methods with deep learning. Sentiment analysis models such as Stock Senti WordNet (SSWN) and Hybrid Quantum Neural Network (HQNN) integrate social media sentiment to capture market dynamics. Real-time frameworks that use stream processing show promise for high-frequency trading applications. This review addresses key challenges including data noise, nonstationarity, overfitting risks, and black-box model interpretability. Solutions include GAN-based synthetic data generation, transformer-based architectures such as SpectralGPT, and optimization techniques for computational efficiency. This review provides a taxonomy of AI-based approaches, while identifying gaps for future research. These findings highlight the potential of AI in financial forecasting while emphasizing the need for interdisciplinary collaboration to address its limitations in data quality, methodology, interpretability, and ethics.