Hybrid Deep Reinforcement Learning for Pairs Trading

Pairs trading is an investment strategy that exploits the short-term price difference (spread) between two co-moving stocks. Recently, pairs methods based on deep reinforcement learning have yielded promising results. These can be classified into approaches: (1) indirectly determining actions and stop-loss boundaries (2) directly spread. In former approach, boundary completely dependent boundary, which certainly not optimal. latter there a risk of significant loss because absence boundary. To overcome disadvantages approaches, we propose hybrid method for called HDRL-Trader, employs independent networks; one other boundaries. Furthermore, HDRL-Trader incorporates novel techniques, such as dimensionality reduction, clustering, regression, behavior cloning, prioritized experience replay, dynamic delay, its architecture. The performance compared with state-of-the-art (P-DDQN, PTDQN, P-Trader). experimental results twenty stock in Standard & Poor’s 500 index show achieves average return rate 82.4%, 25.7%P higher than second-best method, yields significantly positive rates all pairs.