Sentiment Analysis Empowered Drug Recommendations

The process of gathering, detecting, and monitoring information regarding side effects, adverse reactions, and precautions of pharmaceutical products poses significant challenges. With the emergence of user forums, online reviews have become a crucial source of product information. The outbreak of flu, which has led to fatalities, has caused disruption in the medical community, prompting individuals to self-medicate due to limited access to medical consultation, exacerbating their health issues. Recent advancements in machine learning and automation have shown promise across various applications. This study aims to propose a strategy for prescribing medications that can alleviate the workload of specialists. A medication recommendation system is developed in this project, utilizing techniques such as Bag of Words (Bow), TF-IDF, Word2Vec, and manual feature analysis, along with patient ratings to predict sentiment. Various classification algorithms are employed to suggest the most suitable medication for a specific ailment. The sentiment analysis of drug reviews is studied using machine learning classifiers, including Logistic Regression, Perceptron, Multinomial Naive Bayes, Decision Tree, Random Forest, LGBM, and CatBoost applied to Word2Vec, as well as the manual features approach. Additionally, classifiers utilizing Bow, TF-IDF, and Bow using SVC and stochastic gradient descent are utilized. The evaluation metrics include f1-score accuracy, precision, recall, and AUC score. The data-set used in the drug recommendation system based on sentiment analysis of drug reviews comprises drug names, drug sentiment, and drug ratings.