Integrating Textual and Financial Time Series Data for Enhanced Forecasting

Time series forecasting is crucial in various real-world dynamic systems, with machine learning techniques increasingly applied for this purpose. Despite extensive research, there is still a lack of specialized deep learning-based forecasting models in the financial domain. Financial time series, su...

Full description

Saved in:
Bibliographic Details
Published in2024 16th IIAI International Congress on Advanced Applied Informatics (IIAI-AAI) pp. 541 - 544
Main Authors Mou, Shangyang, Xue, Qiang, Zhang, Wenting, Kinkyo, Takuji, Hamori, Shigeyuki, Chen, Jinhui, Takiguchi, Tetsuya, Ariki, Yasuo
Format Conference Proceeding
LanguageEnglish
Published IEEE 06.07.2024
Subjects
Accuracy
Biological system modeling
Economics
Feature extraction
Financial Analysis
Financial Time Series Forecasting
Informatics
Machine learning
Machine Learning in Economics
Numerical models
Predictive models
Real-time systems
Time series analysis
Transformer Model
Online AccessGet full text
DOI10.1109/IIAI-AAI63651.2024.00103

Cover

More Information
Summary:Time series forecasting is crucial in various real-world dynamic systems, with machine learning techniques increasingly applied for this purpose. Despite extensive research, there is still a lack of specialized deep learning-based forecasting models in the financial domain. Financial time series, such as stock prices and gold prices, present significant predictive challenges due to their complex short-term and long-term temporal dependencies. Furthermore, they are influenced by underlying fundamentals often reflected in diverse news sources. Current financial time series prediction methods typically concentrate on numerical data, often overlooking the potential insights offered by textual information. Even when considering textual data, the prevailing approach is to utilize sentiment information extracted from news sources as features. This paper introduces an innovative methodology that integrates relevant financial news content directly with corresponding time series data, aiming to explore the synergistic effects of numerical and textual information to enhance the accuracy of financial time series forecasting. Our experimental results demonstrate a substantial improvement in predictive accuracy when considering related textual data, offering a promising avenue for more nuanced and comprehensive economic market analysis.
DOI:10.1109/IIAI-AAI63651.2024.00103