Improving Drug Review Categorization Using Sentiment Analysis and Machine Learning

Content created by users from social media platforms has become increasingly important as the internet has grown, and it now contains a plethora of knowledge regarding medications, diagnostics, treatments, and illnesses. Customers' opinions regarding previously used medicines are contained in the data in the form of comments, which can be used to detect crucial adverse drug reactions. Using SA techniques like SA, this information can be used to get important insights. It is usually impossible for a prospective buyer to read through all of the comments prior to making a purchasing decision. Drug evaluations benefit both healthcare providers and the general population since they provide vital healthcare data. Assessing opinions regarding many aspects of drug evaluations can provide substantial insights, facilitate choice-making, and improve public monitoring systems by revealing collective perspective. Another key issue is the unstructured and linguistic nature of the evaluations, which finds it challenging for users to categorize comments into useful insights. Previous research has used ML algorithms to perform categorization on drug reviews. So, using appropriate NLP and ML algorithms, our key objective in this study is to acquire a higher categorization score than earlier research studies. We achieved our aim by applying SA on medicine reviews to detect positive, negative, and neutral user comments training five ML algorithms on TF-IDF and CV feature extraction technique. Our results show that out of all ML algorithms, RF trained on CV surpass previous study results, with accuracy and F1 score of 96.65% and 96.42% respectively.