This study introduces a machine learning recommendation system developed to predict disease and present appropriate treatments, consisting of both traditional and alternative medicine approaches. The system uses two organized boosts and random forests to analyze custom symptoms to provide more accurate and personalized medical recommendations. This model was developed with Python microservices and integrated into Gemini-API. Ensures data protection for user data and improves the relevance of proposals. Comparative performance analysis shows that the proposed system exceeds existing models (System A and System B) across several metrics, achieving 90% accuracy, 91% recall, and 91% F1 scores. Furthermore, expert validation values and user satisfaction highlight the reliability and validity of the 83% or 74% model. Despite certain limitations such as potential algorithm distortions in B and external dependence on Gemini-API, the system implies a critical step into AI-controlled, patient-oriented health care that evaluates both traditional and alternative therapies.