arjun-sengupta / twitter-data-sentiment-analysis

Mining Social Network Data for Predictive Personality Modelling by Employing Machine Learning Techniques

Mining Social Network Data for Predictive Personality Modelling by Employing Machine Learning Techniques

Facebook, Twitter, LinkedIn and Tumblr are online social networking platforms where the users send and receive messages on the topic of their choice and express their sentiments. The usage of these sites has exponentially increased over the last few years, thereby increasing the information posted on online social media sites. The quantity of information/tweets keeps increasing on a daily basis. Twitter has become a stable platform to identify personality-related indicators and encrypted in user profiles and pages related to a subject. In this proposed work, we present a scalable real-time system for sentiment analysis of Twitter data. This work will collect tweets of the users in real time and thus provide a basis to identify each tweet into either positive or negative based on the mind-set of the user, thereby providing a real-time analysis of the users regarding a certain topic. The system relies on feature extraction from the tweets generated in real time. A supervised learning approach based on ensemble learning is used to train various classifiers based on the features extracted. A design and implementation in Flask and Celery has been carried out which contains the feature extraction and classification tasks. The system is scalable with respect to the size of the input data and the rate of data arrival. The merits of the proposed system in terms of scalability, performance and classification accuracy was evaluated experimentally.