AI - Powered Medical Chatbot for Symptom Check

Abstract

This paper discusses the design and development of an AI-powered medical chatbot that acts as an intelligent symptom checker and initial healthcare advisor. The system uses Natural Language Processing (NLP) to preprocess user input through tokenization, stemming, and Bag-of-Words (BoW) vectorization, converting unstructured text into a machine-readable format. It employs a Decision Tree Classifier and K-Nearest Neighbors (KNN) model trained on a dataset containing over 130 symptoms and more than 40 diseases to accurately predict the likely disease based on user-reported or selected symptoms. Additionally, it provides precautionary measures, medicine recommendations, and context-aware suggestions for nearby doctors using local datasets. All interactions are securely stored in a MySQL database, allowing users to track their medical history over time. The chatbot operates on a Flask backend that integrates the trained machine learning models, ensuring real-time response generation and smooth data flow from input to prediction. Experimental results demonstrate a 94.2% accuracy with minimal overfitting, validating the model’s reliability and scalability. This system offers an affordable, accessible, and user-friendly digital healthcare solution, particularly beneficial for early disease detection and timely consultation in remote or resource-limited areas, thereby reducing the burden on healthcare professionals.

Ähnliche Arbeiten