Sentiment Analysis and Demand Forecasting from Patient Reviews

The purpose of this paper is to offer pharmaceutical companies and physicians an opportunity to explore the experience of medicines. The challenge though is how to analyze the huge amount of unstructured text to be able to arrive at clear and useful insights. A solution in the present research is proposed which examines large volumes of reviews on medicines through natural language processing (NLP) and machine learning. A system has been designed to clean and process the text then extract features for analysis. In an attempt to quantify the sentiment of the patients, several classification models have been explored and the LinearSVC model is the most accurate and it succeeds in classifying sentiments with an accuracy of more than 94%.