Inteligência artificial para os mercados financeiros: inovações, desafios e oportunidade

Abstract

THE INCREASING VOLATILITY OF FINANCIAL MARKETS AND THE MASSIVE VOLUME OF AVAILABLE DATA HAVE EXPOSED THE LIMITATIONS OF TRADITIONAL STATISTICAL MODELS IN PRICE FORECASTING. THIS WORK CONDUCTS A SYSTEMATIC REVIEW OF THE USE OF ARTIFICIAL INTELLIGENCE (AI) IN FINANCIAL FORECASTING, COMPARING ITS PERFORMANCE WITH CLASSICAL APPROACHES. THE METHODOLOGY WAS BASED ON THE PICO STRATEGY AND THE PRISMA PROTOCOL, RESULTING IN THE ANALYSIS OF 83 ARTICLES. THE RESULTS SHOW THAT MODELS SUCH AS NEURAL NETWORKS, SVM, LSTM, AND HYBRID TECHNIQUES OUTPERFORM TRADITIONAL METHODS IN METRICS SUCH AS MAPE, RMSE, AND SHARPE. HOWEVER, GAPS IN INTERPRETABILITY AND PRACTICAL APPLICABILITY ARE IDENTIFIED. THE REVIEW POINTS TO PATHWAYS FOR FUTURE RESEARCH IN THE FIELD.