An End-to-End Deep Learning Model for Clustering-based Statistical Arbitrage_Binance Futures_67 cryptocurrencies_logreturns

We propose an end-to-end deep learning framework for statistical arbitrage in the cryptocurrency futures markets, which integrates clustering and trading within a unified structure. Statistical arbitrage seeks to exploit price deviations among similar assets to generate market-neutral profits. The cryptocurrency markets, characterized by structural similarities among assets and market inefficiency, offer a promising profitability for such strategies. We incorporate clustering into the model to learn optimal arbitrage trading strategies that capture dynamic co-movements and detect arbitrage opportunities. A transformer encoder is employed to model temporal dependencies among assets. The entire arbitrage process \u2014 from portfolio selection to trade allocation \u2014 is jointly optimized under a custom objective function that reflects the core principles of traditional statistical arbitrage. Empirical results demonstrate that the proposed model consistently outperforms benchmark strategies, achieving strong returns even after accounting for transaction costs. Sub-period analysis shows that the strategy remains robust across varying market conditions, supporting its market-neutral property. Ablation studies further validate the crucial contribution of each model component - clustering, the end-to-end structure, and the proposed objective function- in enabling effective statistical arbitrage.