A recent study reveals a new approach to improve stock market fluctuation predictions using persistent homology. This approach improves the accuracy of various forecast models and marks a significant advance in the integration of topology and finance.
In a new study published in The Journal of Finance and Data Science, researchers from the International Business School of HAN University of Applied Sciences in the Netherlands introduce the theory of topological tail dependence - a new method for predicting stock market fluctuations in turbulent times.
"This research creates a bridge between the abstract world of topology and the practical world of finance," said Hugo Gobato Souto, the study's sole author. "What's really exciting is that this merger gives us a powerful tool that allows us to better understand and predict stock market behavior during turbulent times."
The difference between the average distance of normalized stock returns in two different periods can be used as an indicator for predicting periods of financial turbulence by defining a threshold to be used in normal periods, since the average distance in normal periods is higher than in previous and turbulent periods. However, the problem with this approach is that the average distance of normalized stock returns suffers from the curse of dimensionality and cannot discover nonlinear and complex relationships in the data. The reason the average distance of normalized stock returns suffers from the curse of dimensionality is that as the number of dimensions (or stocks in this case) goes to infinity, the ratio of the distance from any point (say A and B) to the distance from any other point (say A and C) approaches 1. Therefore, the average distance becomes meaningless. On the other hand, implementing pH information through WD or L^n specifications of persistent landscapes does not suffer from these problems. Therefore, this is also the reason why pH information has been successfully used in recent studies and why pH information was chosen for this study. The picture above is a three-dimensional scatter plot from December 16, 2019 to January 16, 2020 (normal period).
Enhancing financial forecasting with persistent homology
Through empirical testing, Souto proved that the addition of persistent homology (PH) information can significantly improve the accuracy of nonlinear models and neural network models in predicting stock market fluctuations in turbulent times.
Soto added: "These findings mark a major shift in the field of financial forecasting, providing investors, financial institutions and economists with more reliable tools."
Notably, this method circumvents the dimensionality barrier and is therefore particularly suitable for detecting complex correlations and nonlinear patterns that are often undetectable by traditional methods.
"It is fascinating to observe the continued improvement in forecast accuracy, especially during the 2020 crisis," Suto said.
Broad implications and future directions
These findings are not limited to one specific type of model. It spans a variety of models, from linear to non-linear models and even advanced neural network models. These findings open the door to improvements in financial forecasting across the board.
Soto concluded: "These findings confirm the validity of this theory and encourage the scientific community to delve deeper into this exciting new intersection of mathematics and finance."
Reference "Topological Tail Dependence: Evidence for Forecasting Realized Volatility," by Hugo Gobato Souto, October 14, 2023, Journal of Finance and Data Science.
DOI:10.1016/j.jfds.2023.100107
Compiled source: ScitechDaily