Healthswitch: A Personalized Medicine Recommender System

ABSTRACT HealthSwitch addresses the critical challenge of finding context-aware alternative medicines by providing a machine learning-powered web application that recommends substitutes aligning in both therapeutic purpose and pharmaceutical form. At its core, the system employs a novel hybrid recommendation engine that synergistically combines the lexical matching capabilities of TF-IDF with the deep semantic understanding of Sentence-BERT embeddings, optimized via grid search to heavily weigh a medicine's form alongside its intended use. Supporting this algorithm is a comprehensive MLOps framework integrated into an administrative dashboard, which provides real-time monitoring of performance trends, hyperparameter drift, and diagnostic visualizations like confusion matrices, facilitating a continuous improvement loop through seamless model retraining and analysis of user feedback. This integration of a sophisticated, fine-tuned recommendation algorithm with robust operational capabilities results in a practical and maintainable end-to-end solution for the complex problem of medicine substitution. KEYWORDS Medicine Recommender System, Hybrid Model, Sentence-BERT(SBERT), TF-IDF, Natural Language Processing, Semantic Similarity, MLOps, Model Monitoring, Vector Embeddings, Drug Substitution