Abstract: Financial market series are significantly influenced by noise, non-stationarity, and unexpected events, posing challenges to stock trend prediction in terms of feature representation and generalization stability. Existing methods predominantly rely on single technical indicators or price series, which inadequately capture the impact of investor behavior on market fluctuations, and their predictive performance and stability remain limited. To address these issues, this paper proposes an attention-driven integrated time-series framework named Blending-ATT-Ridge. First, feature engineering is applied to construct and select multiple financial technical indicators. Then, BiLSTM-ATT and GRU-ATT are employed to model temporal dependencies separately and generate prediction components. Finally, Ridge regularization-based weighted fusion is adopted to integrate these components and mitigate overfitting. Experimental results on real stock datasets from leading Hong Kong-listed enterprises demonstrate that Blending-ATT-Ridge significantly outperforms other models across metrics such as MAE, MAPE, and RMSE. Furthermore, it maintains stable generalization capability under different market conditions, validating the effectiveness and robustness of multi-stage feature engineering and ensemble strategies in stock trend prediction.