Time series is not enough: Financial Transformer Reinforcement Learning for portfolio management

Open Webpage

Xiaotian Ren, Ruoyu Sun, Zhengyong Jiang, Angelos Stefanidis, Hongbin Liu, Jionglong Su

Published: 2025, Last Modified: 04 Jan 2026Neurocomputing 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•We propose FTRL, a novel DRL framework built upon a Financial Transformer backbone.•FTRL jointly models temporal dynamics and latent inter-asset linkages for better decisions.•FTRL outperforms agents based on LSTM and CNN in both return and training efficiency.•It generalizes well across datasets and market regimes, enabling robust portfolio control.