Drug Recommendation system based on Medical Condition Classification and Sentiment Analysis of Drug Reviews

Abstract

The steady growth of the internet has increased the amount of user generated data on the web. In the healthcare domain, patients now commonly post their reviews about medicines after consuming them to create public awareness. Natural Language Processing techniques significantly c ontribute t o the medical field by analyzing these public reviews and identifing the effectiveness of drugs as well as understanding medical conditions they are suffering from which will help healthcare professionals and pharmacovigilance systems to ensure the physical and mental well being of the patients. Hence, this research endeavors to develop a comprehensive framework for patient medical condition classification, s entiment p rediction f rom p atients reviews and recommend suitable medicines to them. Four algorithms: Multinomial Na¨ıve Bayes, Passive Aggressive Classifier, SGD Classifier a nd M LP C lassifier ha ve be en ap plied to medical condition classification a nd t wo a lgorithms: M ultinomial Na¨ıve Bayes and Logistic Regression have been applied for sentiment prediction. The results demonstrate that the proposed framework has an accuracy of 94.4% for Passive Aggressive Classifier in medical condition classification a nd a ccuracy o f 9 4.85% for Logistic Regression in sentiment prediction.