Medinsights: Twitter Based Platform for Health Care Analytics

Twitter is a social media platform where the tweets convey opinions, but the interpretation of this unstructured data can be very time-consuming. Medinsights is a data science platform which leverages machine learning and natural language processing technologies on twitter data. This platform helps to analyze the inherent value of the data extracted from tweets such as diseases, treatments, symptoms linked to healthcare domain, which can only be fully exploited through deep data analytics. Medinsights takes particular health-related query on medicine, disease, brand, medical hashtags as input and converts corresponding twitter data into actionable information in order to gain insights and provide diverse recommendations. Medinsights analyses the sentiment, trend, prevalent location to provide meaningful inferences. Medinsights extracts and proposes associated symptoms, diseases, treatments, drugs and brands based on user query from tweets. For this study gradient boosting classifier is used for tweets classification into medical and non medical domain. Word2vec word embeddings with feed forward neural network is used for sentiment analysis. Conditional random field (CRF) is used to extract medical entities from tweets. Medinsights can be helpful to research correlation from tweets and analytics related to healthcare domain.

I. Introduction

Twitter currently has reached 336 million average active users every month [1]. It is the biggest live open data source for a variety of domain, some part of which has obvious clinical data for the healthcare industry. The value of data is grasped only when this raw information is converted into the knowledge that helps make decisions. Using data science strategy, healthcare organizations can benefit on increasing volumes of data and medical knowledge in an organized, strategic way. Also, individual clinicians can use that knowledge to improve the safety, quality, and efficiency of the care they provide. This project aims to analyze Twitter data and extract various informations like sentiment, prevalent location, trend, diseases, treatments, symptoms, etc. Also, it applies various natural language processing & machine learning technologies such as classification, entity recognition, etc. to gain insights into data related with healthcare domain.