Visual Analytics for Understanding Large-Scale Data Trends
Naisarg Halvadiya
Department of Computer Science
Faculty Supervisor: Shahrukh Humayoun
In today’s data-driven era, extracting insights from vast social media datasets poses significant challenges. This project presents an interactive visual analytics tool that transforms large-scale, unstructured data into intuitive visual representations. Leveraging advanced text processing and sentiment analysis methods, the tool categorizes emotions into eight types (anticipation, trust, surprise, joy, anger, fear, sadness, disgust) alongside basic sentiment polarities. Using a dataset of over 574,000 anonymized English tweets collected during the COVID-19 pandemic from U.S. geolocations, the platform provides multiple interactive views. These include a geolocation-based view mapping emotional trends, a comparative analysis between regions, and a timeline view tracking sentiment evolution. Developed with Python, React, D3.js, and Node.js, the system offers scalable, dynamic insights for researchers, businesses, and policymakers to understand public sentiment and behavioral trends effectively. This innovative solution empowers decision-makers to respond swiftly to emerging social phenomena and guides strategic actions across various sectors. It drives positive change.