Conventional intelligent big data anomaly early warning algorithm for mental health diagnosis and treatment mainly uses MAS (Multi Agent System) multi intelligent distributed centers for early warning assumptions, which is vulnerable to the interaction of early warning, resulting in low early warning consistency. Therefore, this paper designs a new intelligent big data anomaly early warning algorithm for mental health diagnosis and treatment. The research aims to establish an intelligent framework. It involves designing an intelligent model for preemptive detection of irregularities and identifying a track matrix for such data. By achieving this, the research enables intelligent preemptive detection of irregularities in psychological well-being diagnosis and treatment big data. The experipsychological results demonstrate a high level of consistency in the designed algorithm, indicating its strong performance, effectiveness, and practical value. These findings contribute significantly to improving the efficiency diagnosis and treatment.