QuanSIRA: The Quantitative Investment Risk Modeling in Stock Markets with Large Language Models
Abstract: Stock market analysis is important for investors to make financial decisions. Stock price prediction is widely investigated in the natural language processing area due to the superiority of large language models. Recent works have developed several datasets for stock price predictions. However, investment risk, considered an essential factor for investors, is rarely discussed in NLP applications, and there are limited datasets for investment risk analysis. In this work, we propose methods to quantify investment risk and introduce the dataset QuanSIRA. Using this benchmark, we investigate the applications of large language models in tackling quantitative investment risk analysis. The experimental results show the difficulty of investment risk analysis. The model built on pre-trained large language models obtained F1 scores of 68.07 and 65.01 in the in-stock benchmark and the cross-stock benchmark of investment risk prediction task.
Paper Type: Long
Research Area: NLP Applications
Research Area Keywords: financial/business NLP; quantitative investment risk analysis; stock market prediction; pre-trained language model
Contribution Types: NLP engineering experiment, Data resources
Languages Studied: English
Submission Number: 5338
Loading