A review of artificial intelligence quality in forecasting asset prices

Researchers and practitioners globally, from a range of perspectives, acknowledge the difficulty in determining the value of a financial asset. This subject is of utmost importance due to the numerous participants involved and its impact on enhancing market structure, function, and efficiency. This paper conducts a comprehensive review of the academic literature to provide insights into the reasoning behind certain conventions adopted in financial value estimation, including the implementation of preprocessing techniques, the selection of relevant inputs, and the assessment of the performance of computational models in predicting asset prices over time. Our analysis, based on 109 studies sourced from 10 databases, reveals that daily forecasts have achieved average error rates of less than 1.5%, while monthly data only attain this level in optimal circumstances. We also discuss the utilization of tools and the integration of hybrid models. Finally, we highlight compelling gaps in the literature that provide avenues for further research.