Jeremy Moon
Research Mentor(s): Majdi Radaideh
Mentor Department: Nuclear Engineering and Radiological Sciences
Authors: Jeremy Moon, Majdi Radaideh
Session: Session 5 (2:00pm – 2:50pm)
Presentation Type: Poster 36
Abstract
This project presents a comprehensive approach to sentiment analysis of nuclear-related social media posts using machine learning techniques. An automated web scraping solution was implemented, which extracted 120,000+ posts on Instagram Threads—originating from the United States—from 83 nuclear-related keywords using Playwright, a Python automation library. Data preprocessing involved regex-based cleaning and tokenization using various large language models (LLM), such as GPT-2 and BERT. The preprocessed data was then fed into pre-trained machine learning models. A multi-class sentiment classifier was built and trained according to an 80-20 test split on a random sample of 20,000 posts, achieving a 92% accuracy using the Random Forest (RF) and Support Vector Classifier (SVC) models. Finally, an interactive data dashboard to display real-time sentiment statistics was developed by integrating an automated data pipeline that dynamically updates the displayed statistics based on newly collected social media interactions. This study highlights the strengths of the approach, including the collection of relevant metadata—such as the date, likes, comments, and reposts—the potential for replication on other platforms, and a high accuracy of sentiment analysis. However, some limitations are to be noted, such as search function constraints and long processing times. Ultimately, this research contributes to the field of sentiment analysis in the context of nuclear-related social media discourse, providing insights into public opinion and sentiment trends. As a result, policymakers and government entities can use this data-driven approach to monitor public sentiment on nuclear energy, enabling more informed decision-making and proactive engagement with public concerns.
