Multi-View Modeling for Stock Investment Risk Forecasting

TMLR Paper7679 Authors

Abstract: Forecasting stock investment risk is crucial for effective financial decision-making. Existing research on stock risk forecasting is still limited due to the lack of large-scale datasets and standardized investment risk forecasting tasks. To address this problem, we construct a stock investment risk dataset that standardizes the stock risk forecasting task as regression and classification problem, providing a benchmark for stock investment risk forecasting. Recent works only based on time series data capture a limited aspect of historical stock price data. To address this issue, we propose a multi-view framework that leverages large language models (LLMs) and pre-trained vision models to extract complementary long-periodic patterns and short-periodic patterns from historical stock data. Experimental results on our dataset demonstrate that our proposed model outperform the competitive baselines in regression task and classification task of stock investment risk forecasting. The codes and dataset are release in https://anonymous.4open.science/r/MultiV-RF-F87F.

Submission Type: Regular submission (no more than 12 pages of main content)

Changes Since Last Submission:

Since the previous submission, we have made the following revisions:

1、We have updated Table 1, Table 2, Table 3, Table 4, and Figure 4 to enhance clarity and better reflect the experimental results.

2、We have incorporated additional analyses and discussions to strengthen the empirical evidence supporting our method. We add section 6.4 Prompt Strategies and Backbone Models, Appendix A.3 Prompt strategies, A.4 Justification of Dataset Selection and A.5 Normalization Process.

3、We have updated 3.2 Risk Indicator, 5.1 Setting, 5.2 Results, 6.3 Stock Features and Historical Windows and 7 Conclusion.

The revised parts have been highlighted in blue.

Assigned Action Editor: ~Andreas_Lehrmann1

Submission Number: 7679