PaperlayerView exact revision on arXiv

Source-linked AI summary

AI in Finance: Challenges, Techniques and Opportunities

Longbing Cao

arXiv:2107.09051v1q-fin.CPcs.AIcs.CEcs.LG

TL;DR

AI-in-finance research has broad applications but remains divided by disciplinary preferences and research gaps. This review provides a comprehensive roadmap of financial businesses, data, AI techniques, and future opportunities. It concludes that smart EcoFin development must address both AI-enabled innovation and social, ethical, regulatory, privacy, security, transparency, and trustworthiness concerns.

  • Problem

    AI and finance communities pursue different objectives and disciplinary preferences, producing practice differences and research gaps in AI research in finance.

  • Method

    The review categorizes financial businesses and data, surveys classic and modern AIDS techniques, analyzes data-driven methods, and discusses open issues and opportunities.

  • Results

    The review presents a comprehensive overview of classic and modern AIDS techniques, data-driven financial applications, and open opportunities for future AI-empowered finance.

  • Takeaways & Limitations

    Future smart EcoFin research spans AIDS-empowered EcoFin, EcoFin-driven AIDS, and opportunities beyond current AIDS technologies.

  • Takeaways & Limitations

    AI research in finance often treats finance mainly as an application domain, with limited EcoFin theory involvement and domain-friendly interpretation.

Abstract

from arXiv · show

AI in finance broadly refers to the applications of AI techniques in financial businesses. This area has been lasting for decades with both classic and modern AI techniques applied to increasingly broader areas of finance, economy and society. In contrast to either discussing the problems, aspects and opportunities of finance that have benefited from specific AI techniques and in particular some new-generation AI and data science (AIDS) areas or reviewing the progress of applying specific techniques to resolving certain financial problems, this review offers a comprehensive and dense roadmap of the overwhelming challenges, techniques and opportunities of AI research in finance over the past decades. The landscapes and challenges of financial businesses and data are firstly outlined, followed by a comprehensive categorization and a dense overview of the decades of AI research in finance. We then structure and illustrate the data-driven analytics and learning of financial businesses and data. The comparison, criticism and discussion of classic vs. modern AI techniques for finance are followed. Lastly, open issues and opportunities address future AI-empowered finance and finance-motivated AI research.

1 INTRODUCTION

The review situates AI in finance within decades of classic and modern AIDS research, then frames smart FinTech as a synthesis of techniques and EcoFin businesses. It contrasts this broad roadmap with reviews centered on individual techniques or business problems.

  • New-generation AIDS advancement is transforming and synthesizing financial services, the economy, technology, media, communication, and society.
  • Smart FinTech connects AIDS techniques with EcoFin businesses to intelligentize core businesses, operations, services, and decision-making.
  • The smart FinTech family spans BankingTech, LendTech, WealthTech, TradeTech, PayTech, InsurTech, RiskTech, and RegTech.
  • Existing reviews commonly focus on a specific AI technique or a specific financial business problem.
  • The review summarizes EcoFin businesses and their challenges as targets that can benefit from AIDS techniques.

2 AI-EMPOWERED FINANCIAL BUSINESSES AND CHALLENGES

AI-empowered finance spans broad business areas, products, procedures, and system-wide perspectives, all producing extensive entities, interactions, activities, and data. These businesses create opportunities for mechanism design, forecasting, optimization, risk management, and related AI research.

  • AIDS-enabled EcoFin and FinTech encompass nearly all aspects of EcoFin systems, their environments, and businesses.
  • The business landscape includes financial products and services, procedural aspects of ecosystems and markets, and systematic cross-aspect alternatives.
  • These areas involve large quantities of objects, entities, interactions, activities, and lifetime data.
  • AI opportunities include mechanism design, forecasting and prediction, portfolio planning and optimization, sales and marketing analysis, profiling, and risk management.

3 ECONOMIC-FINANCIAL DATA AND CHALLENGES

EcoFin AI relies on diverse internal and external data sources whose characteristics and coupling create research challenges. The review organizes data-driven AIDS around five major analytical families and links them to smart FinTech opportunities.

  • EcoFin data include transactions, macroeconomic indicators, client and operational data, events, news, reports, social media, cognitive data, and accounting-related data.
  • EcoFin businesses and data are coupled in reality, creating challenges involving innovation and business complexity.
  • The review focuses on data-driven AIDS techniques after discussing EcoFin businesses, data, and their challenges.

4 AN OVERVIEW OF AI RESEARCH IN FINANCE

AI research in finance combines mathematical, complex-system, classic analytical, computational-intelligence, and modern learning methods. These techniques analyze diverse financial data and address patterns, uncertainty, temporal dynamics, interactions, and system complexity.

  • The technical family comprises mathematical and statistical modeling, complex-system methods, classic analysis and learning, computational intelligence, and modern analytics and learning.
  • Mathematical and Statistical Modeling: Mathematical and statistical methods quantify and analyze EcoFin systems through numerical, time-series, signal-analysis, and statistical-learning techniques.
  • Complex System Methods: Complexity science, game theory, agent-based modeling, and network science model mechanisms, interactions, evolution, and connections in EcoFin systems.
  • Classic Analysis and Learning Methods: Classic analytics and learning discover or optimize patterns, clusters, classes, trends, outliers, events, behaviors, and document or network relations.
  • Data-Driven Analytics and Learning: Data-driven AI addresses EcoFin data through time series, text, behavior and events, multisource analysis, and deep financial modeling.

5 DATA-DRIVEN AI IN FINANCE

Data-driven AI in finance analyzes heterogeneous economic-financial data, including time series, text, behaviors, events, and multisource inputs. The section surveys analytical objectives and techniques spanning representation learning, sequence modeling, text analysis, behavior modeling, and distributed financial modeling.

  • Financial time-series analysis: Financial time-series analysis covers representation learning, forecasting, and modeling dependencies, correlations, causality, and other interactions.Approaches range from regression, autocorrelation, and signal processing to deep autoregressive models, LSTMs, and CNNs.
  • Economic-financial text analysis: Economic-financial text analysis extracts fraud, events, knowledge, sentiment, topics, opinions, and company-, instrument-, product-, or service-related information.Methods include document analysis, NLP, topic modeling, sentiment analysis, embeddings, recurrent models, Transformers, and attention networks.
  • Behavior and event modeling: Behavior and event modeling examines investment transactions, behavioral sequences, performance patterns, exceptional activity, and differences among groups.Sequence models, event detection, multitask learning, point processes, Hawkes processes, and Markov processes support these analyses.
  • Multisource financial data analysis: Comprehensive financial analysis often requires multisource data combining numerical, textual, visual, tabular, temporal, sequential, and cross-market information.These inputs support modeling relationships among indicators, integrating fundamental and technical analysis, assessing event effects, and predicting prices or portfolios.
  • Deep financial modeling: Deep distributed financial modeling is an increasingly important AI-in-finance area alongside deep representation, prediction, cross-market, sector, and factor modeling.Typical techniques include distributed machine learning, deep transfer learning, federated learning, and cloud analytics.

6 GAPS IN THE AI RESEARCH IN FINANCE

AI and finance communities pursue different objectives and evaluation practices, creating interdisciplinary gaps in AI research in finance. The paper compares major AI technique families and organizes opportunities for more integrated, finance-aware development.

  • 6.1 Difference between AI and Finance Research: AI and finance communities show strong disciplinary preferences and objectives, resulting in practice differences and research gaps.
  • 6.1 Difference between AI and Finance Research: Finance-driven research often uses simple AI methods, small data, and finance-oriented evaluation while giving less attention to systemic AI evaluation.It commonly treats AI as a complementary tool for explaining, simulating, and understanding EcoFin phenomena.
  • 6.1 Difference between AI and Finance Research: AI-driven research often emphasizes novel AI methods or applications, large data, and comprehensive AI evaluation while offering less financial-impact demonstration.Finance is frequently treated as an application domain without deep domain understanding or domain-friendly interpretation.
  • 6.2 Pros and Cons of AI Research in Finance: The review covers six major technique families: mathematical and statistical modeling, complex systems, classic analytics and learning, computational intelligence, modern AIDS, and hybrid AIDS.These families are discussed in terms of their advantages and disadvantages for EcoFin problems and systems.
  • 6.2 Pros and Cons of AI Research in Finance: Increasingly complicated real-life EcoFin problems often require hybridization of complementary classic and modern AI techniques.
  • 6.3 Opportunities: Open opportunities are grouped into AIDS-empowered EcoFin developments, EcoFin-driven AIDS developments, and opportunities beyond current AIDS technologies.The paper frames these opportunities as ways to empower smart EcoFin.

7 OPEN OPPORTUNITIES: SMART FUTURES

The paper identifies smart futures for finance through AIDS-enabled strategic planning, innovation, broader economic-financial applications, and interdisciplinary research. It emphasizes context-aware modeling, distributed learning, and the need to address social, ethical, geopolitical, and economic dimensions.

  • AIDS-driven strategic planning and developments for smart EcoFin: AIDS can support strategic planning by interpreting visions, providing evidence and forecasts, and evaluating plans against historical performance and external practices.Applications include optimizing strategic directions, interpreting decisions, and estimating effects on business performance.
  • AIDS-enabled EcoFin Innovations: Context-aware representations should integrate financial factors, policy constraints, market conditions, macroeconomic events, and changing contextual dynamics when modeling targets such as credit scores or loan defaults.The proposed scope spans individual to compound economic-financial targets and representations of stocks, indices, currencies, companies, and economies.
  • AIDS-enabled EcoFin Innovations: Distributed learning and automated investment systems are opportunities for secure data sharing, privacy protection, feature discovery, portfolio optimization, risk mitigation, and actionable financial analytics.The agenda includes blockchain-based learning, model consensus, automated trading and advising, and discriminative features from technical, fundamental, time-series, and news data.
  • AIDS-empowered strategic and innovative thinking for EcoFin: Smart EcoFin aims to shift finance toward distributed, personalized, human-centric, evidence-based, adaptive, proactive, and personal-demand-driven systems and services.These translations are presented as requiring new-generation AIDS while also motivating further AIDS innovation.
  • Opportunities beyond AIDS: Future research extends beyond mainstream finance to social good, productivity, well-being, security, geopolitical and geocultural adaptation, and the economics of AIDS.The paper also calls for neuroscience-enabled foundations and interdisciplinary study of AIDS's effects on businesses, labor, productivity, jobs, growth, society, and regulation.
  • Social and ethical issues in AIDS-driven EcoFin: Smart EcoFin creates unresolved concerns involving fairness, inequality, misuse, robustness, sustainability, security, privacy, transparency, explainability, trustworthiness, and regulatory compliance.Examples include manipulation of currency rates and intangible assets, as well as the need to mitigate negative effects of intelligence, robotics, and automation.
  • Section overview: The review synthesizes decades of classic and modern AIDS research in finance, critiques their advantages and weaknesses, and identifies open issues and future opportunities.Its scope complements reviews focused on individual techniques or financial problems.

8 CONCLUSIONS

The supplied conclusion material records partial sponsorship from the Australian Research Council.

  • 8 CONCLUSIONS: The work was partially sponsored by two Australian Research Council grants.The grants are Discovery grant DP190101079 and Future Fellowship grant FT190100734.
Loading 2107.09051v1…