Effectiveness of Uncertainty Consideration in Neural-Network-Based Financial Forecasting

Abstract

Accurate prediction of financial markets is considered one of the most difficult problems due to the nature of its complexity, influenceability, and nonstationarity. Recent financial forecasting applications using neural networks typically have not taken the predictive uncertainty into consideration. Without proper consideration of predictive uncertainty, such approaches may lead to unintended investment losses. Therefore, consideration of predictive uncertainty in neural network-based financial forecasting should lead to improved investment decision-making. In this study, the effectiveness of uncertainty consideration in neural network-based financial forecasting was verified through a simulated investment portfolio. We show that ensemble and Bayesian neural network models are effective in realizing more stable investment outcomes.