DRUG RECOMMENDATION SYSTEM BASED ON SENTIMENT ANALYSIS OF DRUG REVIEWS USING PASSIVE AGGRESSIVE CLASSIFIER

Abstract

The COVID-19 pandemic has severely strained healthcare resources, leading to a shortage of specialists, medical equipment, and medicines. As a result, many individuals have resorted to self-medication without proper consultation, often worsening their health conditions. To address this issue, we propose a Drug Recommendation System that utilizes sentiment analysis of patient reviews to assist in selecting the most effective medications. Our approach involves preprocessing drug review data by removing stop words, correcting misspellings, and tokenizing text. We employ TF-IDF vectorization for feature extraction and use the Passive Aggressive Classifier for sentiment classification, predicting whether a review is positive, neutral, or negative. The model is evaluated using accuracy, precision, recall, F1-score, and AUC-ROC, with results indicating that the Passive Aggressive Classifier with TF-IDF provides robust and efficient sentiment classification. Unlike traditional models that merely classify reviews, our system integrates sentiment scores to recommend the most suitable drugs for specific medical conditions. Additionally, Word2Vec-based Exploratory Data Analysis (EDA) is conducted to enhance feature representation and sentiment trends. This research aids both healthcare professionals and patients by offering data-driven medication insights based on real-world reviews. Future work will focus on deep learning integration, user-specific recommendations, and dataset expansion to improve the system’s predictive accuracy and personalization.