Financial Aid Model for AccessUVA

Introduction

• Financial time series forecasting in financial aid faces challenges due to limited data and complex trends.

• Big data analytics and predictive analysis are key approaches for financial forecasting.

• Challenges in Financial Aid (FA) time series analysis:
  • Limited historical datasets
  • High dimensional financial information
  • Balancing accuracy with computational efficiency
• Proposed solution:
  • Pre-trained foundation models
  • State-of-the-art time series models, including:
    • GPT-2 as the backbone LLM
    • Transformer architectures
    • Linear models
• Advantages:
  • Outperforms traditional methods
  • Effective with minimal (few-shot) or no fine-tuning (zero-shot)
• Benchmark study:
  • Covers financial aid and seven other time series tasks
  • Highlights LLMs’ potential for scarce financial datasets

Publications

Large Language Models for Financial Aid in Financial Time-series Forecasting IEEE Big Data International Workshop on Large Language Models for Finance, Washington DC, USA, December 2024