This paper embarks on an extensive exploration of the challenges inherent in traditional three-way decision-making methodologies. First, these conventional approaches necessitate precise subjective threshold definitions, rendering them less adaptable across diverse scenarios. Second, their interpretability in information fusion is often lacking, resulting in ambiguous decision criteria. To address these issues, this paper introduces a novel approach that leverages complex fuzzy sets to establish an adjustable multi-granularity complex fuzzy information system. It integrates decision-based three-way decision techniques to derive classification thresholds objectively, yielding an adjustable multi-granularity complex fuzzy probabilistic rough set. Furthermore, the paper acknowledges the constraints of bounded rationality that decision-makers encounter in real-world scenarios. Within the framework of behavioral decision theory, this work provides a more comprehensive depiction of how psychological factors influence decision outcomes. In essence, this paper proposes a multi-granularity three-way decision model grounded in complex fuzzy sets and regret theory, designed to tackle complex multi-attribute group decision-making problems. To validate the model's effectiveness in addressing decision challenges related to decision risk and bounded rationality within a complex fuzzy environment, the paper conducts an in-depth experimental analysis using a shared bicycle rental dataset. The results underscore the model's feasibility and efficacy, particularly in enhancing fault tolerance and leveraging expert experience.