Innovative Drug Recommendation Systems: Decision Making Through Patient Reviews and Attention Mechanism

Abstract The choice of the right drug is critical to the effective treatment and recovery of a patient. With more patient reviews and feedback available, it is becoming easier to incorporate this data into decision-making systems for personalized drug recommendations. The purpose of this paper is to introduce an integrated patient disease detection and drug recommendation system that incorporates patient reviews to assist in selecting the appropriate medication for individual patients. The proposed system employs a FL-DistilBERT-BiGRU-Attention model, which combines the efficiency of DistilBERT (a distilled, faster variant of BERT) for extracting meaningful language features with the sequential learning capabilities of BiGRU (Bidirectional Gated Recurrent Unit) and the interpretability of an attention mechanism. To improve model performance, especially in cases of class imbalance, Focal Loss (FL) is utilized to emphasize hard-to-classify samples. Drug reviews are used to diagnose a patient’s condition, and a multi-criteria decision-making process is used to recommend effective medication based on user feedback about the effectiveness of these drugs. Data-driven insights based on patient feedback are provided to healthcare professionals and patients to enhance drug recommendation decision-making. This model is effective and computationally efficient in delivering accurate drug recommendations across a variety of disease categories and drug reviews, demonstrating a practical framework for patient-centered health care with a practical and balanced approach.