Fuzzy covering rough set models have shown strong potential in uncertainty modeling and attribute reduction, yet existing neighborhood-based measures often struggle to simultaneously capture local similarity, global interactions, and controllable approximation granularity. To address these issues, this paper proposes a comprehensive $(\alpha ,\beta )$-neighborhood measure framework built upon fuzzy β-covering approximation spaces. The proposed construction introduces two parameters to flexibly regulate the neighborhood formation and approximation strictness, thereby improving robustness in the presence of boundary and uncertain samples. Furthermore, an overlap-based aggregation mechanism is incorporated to characterize object relationships more effectively, while generalized Shapley values are employed to quantify attribute importance under interaction effects. On this basis, a Shapley–Choquet integral decision model is developed for non-additive information fusion, enabling more expressive multi-attribute evaluation than conventional additive weighting schemes. Experimental results on representative datasets demonstrate that the proposed framework achieves stable performance improvements in attribute reduction and decision-making tasks, especially when strong attribute dependencies exist. These findings confirm the effectiveness of integrating fuzzy neighborhood modeling with interaction-aware aggregation for uncertainty-aware data analysis.