Uncertain online portfolio selection with LSTM predictors

Online Portfolio Selection (OLPS) has emerged as a rapidly advancing field at the intersection of financial engineering and artificial intelligence, aimed at maximizing cumulative wealth through sequentially adjusting portfolio allocations in dynamic market environments. The core challenge for online portfolio selection lies in accurately forecasting the prospective yields of volatile assets and deriving best portfolio allocations instantaneously. Traditional approaches often rely on historical return patterns and probabilistic assumptions, which often fail to capture complex temporal dependencies and adequately quantify inherent market uncertainties. To address these limitations, this work studies the OLPS problem under the framework of uncertainty theory and introduces a novel framework that synergistically integrates Long Short-Term Memory (LSTM) networks to generate precise return predictions. Based on this dual-pathway design, the adaptive uncertain mean-absolute deviation optimization model is designed, which dynamically balances uncertainty-adjusted expected return against decomposed risk metrics and transaction costs. Finally, several numerical experiments are conducted and solved to illustrate the efficacy and benefits of the proposed approach.