The findings indicate that machine learning effectively compensates for the defects of traditional methods, yet faces common challenges including weak model interpretability, high dependence on high-quality data, and low cross-market adaptability.
Abstract
With the rapid development of fintech, machine learning has been widely applied in core financial fields, becoming an important technical tool to address practical challenges in financial research and practice. Traditional quantitative and qualitative methods show obvious limitations in handling complex futures price fluctuations, precise valuation of diversified enterprise assets, and dynamic early warning of multi-factor financial risks. Many international studies have explored machine learning applications in futures price forecasting, enterprise asset pricing, and risk management separately, yielding fruitful results. However, a lack of systematic review and integrated analysis across these three areas restricts a comprehensive understanding of machine learning’s overall role in corporate investment and management. This paper reviews recent international research on machine learning in futures price prediction, enterprise asset pricing, and corporate risk control. It summarizes the performance of different machine learning algorithms, compares their methodological features, data processing approaches, and practical effects. The findings indicate that machine learning effectively compensates for the defects of traditional methods, yet faces common challenges including weak model interpretability, high dependence on high-quality data, and low cross-market adaptability. This study fills the gap in systematic reviews of integrated machine learning applications in the above financial fields, improves the relevant research system, offers practical guidance for global investors, enterprises and financial institutions, and points out directions for future cross-field research.
Predicting the stock market has become an important research topic in recent years, because accurate result can support investment decision and reduce financial risks.Traditional statistical models are hard to analyze in the nonlinear market, leading researchers to develop machine learning models. This paper reviews recent studies about stock prediction based on machine-learning methods from four perspectives: traditional machine learning models, sentiment analysis, technical indicators, macroeconomic and fundamental factors. The reviewed literature shows that machine learning methods can have better result than traditional models in handling complex financial data. Furthermore, sentiment information, technical indicators, and macroeconomic or fundamental factors can significantly improve prediction performance. However, each approach has limitations. The review finds that using multiple factors may relate to accurate predictions. Therefore, combining different factors through hybrid models is considered the most effective strategy. Finally, this review reveals a clear transition from single-factor prediction models to multi-factor prediction systems and provides a comprehensive understanding of current developments in stock prediction and highlights potential directions for further research.
Tianyuan Shi· Applied and Computational En...· 0 citations
The paper shows that ensemble learning models have superior predictive power and argues that behavioral data complement traditional datasets for underbanked populations, such as "credit invisibles," and makes a strong case for XAI being essential for model transparency, combating bias, and meeting regulatory requirements.
Bojun Chen· Advances in Economics, Manag...· 0 citations
The importance of predicting the price of gold using artificial intelligence systems is consid-ered. Proofs concerning the importance of accurate forecasting of price trends in the gold indus-try, which is a key factor for investors, financial institutions, and economic analysts, are pro-vided. Using AI in this context can help to improve risk management strategies and decision-making in financial markets. Today, people use intelligent monitoring to obtain information about the properties of an object or process by creating and using a model knowledge base while processing the results of observations. When using intelligent monitoring to forecast financial indicators, there is a need to synthesize forecasting models based on limited information about the process history. Each future value of the forecasted indicator is determined by factors that took place in the past. The modeling of exchange bond pricing processes occurs under both structural and informational uncertainty. In order to reduce the uncertainty of the process of forecasting stock indicators, the paper presents the research results using a new method of ma-chine learning as an additional structural element in combination with already existing algo-rithms for synthesizing models of a multilayer agent synthesizer of predictors. The use of a new element does not always improve the characteristics of the system as a whole. The hypothesis about the improvement of the characteristics of model agent synthesizers when using a new ma-chine learning method as a structural element of the layer was tested. For example, the process of forecasting prices of gold on the stock exchange was studied. Simultaneously with the crea-tion of new methods of machine learning, it is proposed to change the structure of the multi-layer synthesizer of models. It was investigated how the new properties of the structural ele-ment of one of the layers change the structure of the agent synthesizer of models as a whole. The research results prove the effectiveness of the process of building a new method of machine learning as a structural element of a multi-layer agent synthesizer of models.
S. Holub, D. Tolbatov· Mathematical machines and sy...· 0 citations
Stock market prediction has become an important research area due to its significant role in supporting investment decisions, risk management, and financial planning. The highly dynamic and nonlinear nature of financial markets makes accurate forecasting a challenging task, encouraging researchers to adopt Artificial Intelligence-based techniques. In recent years, Machine Learning, Deep Learning, and Hybrid approaches have demonstrated remarkable improvements over conventional statistical methods in predicting stock prices and market trends. This paper presents a comprehensive Systematic Literature Review of recent advancements in AI-based stock market prediction published between 2020 and 2026. The review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses methodology to systematically identify, screen, evaluate, and synthesize relevant peer-reviewed studies collected from major scientific databases. A total of 30 high-quality studies were selected for detailed analysis and categorized into Machine Learning, Deep Learning, and Hybrid approaches. The selected studies were comparatively analyzed based on prediction models, datasets, evaluation metrics, forecasting performance, publication trends, and application domains. The findings indicate that Long Short-Term Memory is the most widely adopted Deep Learning model, XGBoost and Random Forest are the dominant Machine Learning algorithms, while Hybrid models consistently achieve the highest prediction accuracy and lower forecasting errors. The review also identifies key research challenges, including market volatility, non-stationary financial data, model interpretability, and computational complexity, and highlights emerging research directions such as Explainable Artificial Intelligence, multimodal data integration, Transformer-based architectures, Graph Neural Networks, and adaptive learning frameworks. This review provides a comprehensive reference for researchers and practitioners by summarizing recent developments, identifying existing research gaps, and outlining future opportunities for developing more accurate, robust, and intelligent stock market prediction systems.
Sunil Molke, A. Vaidya, G. P. Dhok· International journal of adv...· 0 citations
In the context of the digital economy, consumer demand, platform traffic, and supply chain environments are constantly changing. Accordingly, business forecasting has laid a critical foundation for enterprise inventory scheduling, marketing resource allocation and operational risk management. Compared with traditional statistical forecasting methods, such as ARIMA and exponential smoothing, machine learning can process multi-source, high-dimensional, and nonlinear information through big data and algorithms. It has gradually become an important research direction in business forecasting. This paper adopts systematic literature review and typical case analysis, to compare machine learning with traditional forecasting methods, explore its practical value in sales prediction, customer churn early warning and demand fluctuation evaluation, and sort out corresponding operational, compliance and technical risks as well as standardized governance requirements. The findings show that machine learning can improve forecasting accuracy and dynamic responsiveness while further supporting inventory and supply chain coordination, marketing decision-making, customer segmentation, and personalized recommendation. However, its effectiveness is also constrained by data quality, model interpretability, privacy compliance, and technical security. To solve the above problems, enterprises need to build a complete governance system covering standardized data management, scientific model supervision and manual human oversight. This paper contributes to understanding the transition of business forecasting from the traditional statistical paradigm to the intelligent forecasting paradigm and provides practical guidance for enterprises seeking to balance technological application, operational optimization, and risk control.
Chenxi Qiu· Advances in Economics, Manag...· 0 citations
Profitability deterioration often appears before bankruptcy, default, or formal financial distress. This paper develops an explainable machine learning framework to identify listed companies with high next-year profitability downside risk. Using a public firm-year panel constructed from SEC Financial Statement Data Sets and Stooq historical stock price data, this study examines U.S. non-financial listed companies from 2015 to 2024. High-risk observations are defined as firms whose next-year ROA change falls in the bottom 30 percent within the same industry-year group. Logistic Regression, Random Forest, XGBoost, and LightGBM are evaluated under a chronological validation design. XGBoost performs best in the out-of-sample test set, with an AUC of 0.836 and a PR-AUC of 0.653. SHAP results indicate that revenue growth, operating margin, leverage, ROA, operating cash flow growth, and stock volatility are the main risk drivers.
Xiaoqian Lin· Applied and Computational En...· 0 citations
