🤖 AI Summary
This study addresses the challenge of stock return prediction in non-stationary financial markets by proposing an adaptive Financial Transformer model. The architecture incorporates a market state encoder and an adaptive gating attention mechanism to dynamically integrate eleven semantically grouped financial indicators, thereby modulating temporal modeling according to prevailing market conditions. A finance-aware composite objective function is introduced to jointly optimize prediction error, directional accuracy, and the Sharpe ratio, while correcting sequence alignment bias commonly present in backtesting procedures. Empirical evaluations demonstrate that the proposed method achieves competitive predictive performance across multiple stocks, reduces model complexity by 15.2%, and significantly enhances both parameter efficiency and the interpretability of trading signals.
📝 Abstract
Adaptive Financial Transformer (AFT) is proposed for stock return prediction under non-stationary financial markets. The model incorporates a Market Regime Encoder, an Adaptive Gate Network, and an Adaptive Financial Context module to dynamically bias self-attention based on semantic relationships between financial indicators. Unlike conventional Transformer architectures that treat all input features uniformly, the proposed approach groups 95 engineered financial features into 11 semantic categories and adapts attention according to latent market regimes. The study also identifies and corrects sequence alignment and backtesting issues that can inflate reported trading performance, and introduces a financially-aware composite objective that jointly optimizes prediction error, directional accuracy, and non-overlapping Sharpe ratio. Extensive experiments compare the proposed architecture against classical machine learning models, recurrent neural networks, and Transformer baselines using chronological evaluation, five random seeds, ablation studies, hyperparameter optimization, explainability analysis, and multi-stock validation. Results demonstrate competitive predictive performance while reducing model complexity by 15.2% and improving parameter efficiency through feature selection, providing an interpretable Transformer architecture for financial time-series forecasting.