Generalized Stock Price Prediction for Multiple Stocks Combined with News Fusion
arXiv:2603.19286v1 Announce Type: cross Abstract: Predicting stock prices presents challenges in financial forecasting. While traditional approaches such as ARIMA and RNNs are prevalent, recent developments in Large Language Models (LLMs) offer alternative methodologies. This paper introduces an approach that integrates LLMs with daily financial news for stock price prediction. To address the challenge of processing news data and identifying relevant content, we utilize stock name embeddings within attention mechanisms. Specifically, we encode news articles using a pre-trained LLM and implement three attention-based pooling techniques -- self-attentive, cross-attentive, and position-aware self-attentive pooling -- to filter news based on stock relevance. The filtered news embeddings, combined with historical stock prices, serve as inputs to the prediction model. Unlike prior studies that focus on individual stocks, our method trains a single generalized model applicable across multipl
arXiv:2603.19286v1 Announce Type: cross Abstract: Predicting stock prices presents challenges in financial forecasting. While traditional approaches such as ARIMA and RNNs are prevalent, recent developments in Large Language Models (LLMs) offer alternative methodologies. This paper introduces an approach that integrates LLMs with daily financial news for stock price prediction. To address the challenge of processing news data and identifying relevant content, we utilize stock name embeddings within attention mechanisms. Specifically, we encode news articles using a pre-trained LLM and implement three attention-based pooling techniques -- self-attentive, cross-attentive, and position-aware self-attentive pooling -- to filter news based on stock relevance. The filtered news embeddings, combined with historical stock prices, serve as inputs to the prediction model. Unlike prior studies that focus on individual stocks, our method trains a single generalized model applicable across multiple stocks. Experimental results demonstrate a 7.11% reduction in Mean Absolute Error (MAE) compared to the baseline, indicating the utility of stock name embeddings for news filtering and price forecasting within a generalized framework.
Executive Summary
This article introduces a novel approach to stock price prediction by integrating Large Language Models (LLMs) with daily financial news. The proposed method utilizes stock name embeddings within attention mechanisms to filter news data and identify relevant content. The filtered news embeddings, combined with historical stock prices, serve as inputs to a generalized prediction model applicable across multiple stocks. Experimental results demonstrate a 7.11% reduction in Mean Absolute Error (MAE) compared to the baseline, highlighting the utility of the proposed framework. The study contributes to the field of financial forecasting by offering an alternative methodology to traditional approaches such as ARIMA and RNNs.
Key Points
- ▸ The authors propose a novel approach to stock price prediction by combining LLMs with financial news
- ▸ The method utilizes stock name embeddings within attention mechanisms for news filtering
- ▸ A generalized prediction model is developed for multiple stocks, reducing the need for individual stock models
Merits
Strength in Addressing Data Heterogeneity
The proposed method effectively addresses the challenge of processing news data and identifying relevant content by utilizing stock name embeddings within attention mechanisms.
Improved Performance
The experimental results demonstrate a significant reduction in Mean Absolute Error (MAE) compared to the baseline, indicating the utility of the proposed framework.
Generalizability Across Multiple Stocks
The development of a generalized prediction model is a significant contribution to the field, allowing for stock price prediction across multiple stocks without the need for individual models.
Demerits
Limited Dataset Consideration
The study may be limited by the consideration of a specific dataset, and the generalizability of the results to other datasets remains to be seen.
Dependence on Pre-trained LLMs
The proposed method relies on pre-trained LLMs, which may be subject to biases and limitations, potentially affecting the accuracy of the predictions.
Lack of Interpretability
The use of attention mechanisms and LLMs may make it challenging to interpret the results and understand the decision-making process of the model.
Expert Commentary
The proposed approach demonstrates the potential of combining LLMs with financial news to improve stock price prediction. However, further research is needed to address the limitations mentioned in the study, such as the dependence on pre-trained LLMs and the potential biases in the results. The development of more interpretable models and the consideration of diverse datasets will be essential to further advance the field of financial forecasting.
Recommendations
- ✓ Future studies should investigate the use of alternative data sources, such as social media and economic indicators, to improve the accuracy of financial forecasting.
- ✓ The development of more interpretable models, such as explainable AI techniques, will be essential to understand the decision-making process of the model and identify potential biases.
Sources
Original: arXiv - cs.AI
