Research

Mood Swings The Market: Role of Sentiments in the GameStop Saga     Dr. Yiwei Zhang, UW Madison - Advisor 

Link to paper : Read full paper here

• Analysed over 300,000 Reddit posts to study the impact of retail investor sentiment on GameStop’s stock during the 2021 short squeeze by combining deep learning with large language model-based sentiment analysis.

• Implemented a sentiment extraction pipeline using Gemma-3 (12B) LLM with Chain-of-Thought prompting resulting in 80% accuracy (based on 600 human annotated posts), outperforming baseline tools (VADER, FinBERT, FinGPT)

• Built and trained multistage LSTM-based non autoregressive sequence models to predict intraday GME stock price using Reddit engagement, scores and net sentiment , achieving higher accuracy during “squeeze" phases than post due to herding and information cascades

Debt Dynamics and Well-Being: A Mobile Survey of Low Income Adults Dr. Megan Bea, UW Madison and Dr Mariana Amorim, Washington State University 

• Collecting and managing biweekly smartphone survey data from 410 low-income adults, building one of the first high-frequency datasets on short-term debt fluctuations and their impact on health

• Developing novel survey instruments to track multiple debt sources and reasons for borrowing, enabling richer insights into financial behavior beyond traditional credit measures.

• Advancing research on debt as a social determinant of health by integrating high-frequency debt and health data with administrative annual credit data.