SafeMeds - Drug Interaction Predictor

The identification of harmful drug–drug interactionsis a major challenge in modern healthcare due to the increasinguse of multiple medications by patients. Traditional methods relyon manual checking, static databases, and physician experience,which are time-consuming and may fail to detect complex ornewly emerging interactions. These approaches lack intelligentprediction capabilities and often provide limited transparency indecision-making.This paper presents SafeMeds– an AI-based drug interaction prediction system designed to provide accurate, fast, andreliable detection of harmful drug combinations. The systemintegrates a web-based interface developed using HTML, CSS,and JavaScript, with a backend powered by Python Flaskand SQLite for efficient data management. Advanced MachineLearning and Deep Learning models, including CatBoost andLSTM, are used to analyze drug interaction patterns and predictrisk levels.The system allows users to input drug names along withadditional parameters such as dosage, age, gender, and existing diseases. It then predicts the interaction risk level (High,Moderate, or Low) and provides confidence scores. Additionally,Explainable AI techniques using SHAP are incorporated topresent transparent explanations for each prediction, helpingusers understand the reasoning behind the results.