This paper addresses the limitations of conventional RBAC role-mining algorithms, which often lack meaningful interpretation and do not facilitate effective user-role assignments. Machine learning models also pose risks due to their probabilistic nature and inaccuracies. As an alternative, Hierarchical Team Permission Analysis (HTPA) offers a scalable solution for role mining in large enterprises, providing clear business interpretations and explicit assignment rules. We enhance HTPA to reduce the minimizing criterion by developing a Role Mining pipeline with optimized hyperparameters. This pipeline, which includes Binary Matrix Factorization (BMF) and logistic regression, achieved a 13% reduction in the criterion compared to standard HTPA by reducing false positives, resulting in improved security and role assignment accuracy. The combined strengths of HTPA and BMF offer a more robust and interpretable role-mining solution.