Using self-similarity and renormalization group to analyze time series

TL;DR

This paper implements and tests a Renormalization Group-based algorithm for analyzing and forecasting financial time series, leveraging self-similarity properties to potentially predict market movements.

Contribution

It explicitly codes and evaluates a RG-based algorithm for financial time series forecasting, highlighting its responsiveness to market changes and potential for detecting sharp movements.

Findings

Algorithm shows responsiveness to different input parameters.

Potential to detect sharp market movements.

Current performance needs improvement.

Abstract

An algorithm based on Renormalization Group (RG) to analyze time series forecasting was proposed in cond-mat/0110285. In this paper we explicitly code and test it. We choose in particular some financial time series (stocks, indexes and commodities) with daily data and compute one step ahead forecasts. We then construct some indicators to evaluate performances. The algorithm is supposed to prescribe the future development of the time series by using the self-similarity property intrinsically present in RG approach. This property could be potentially very attractive for the purpose of building winning trading systems. We discuss some relevant points along this direction. Although current performances have to be improved the algorithm seems quite reactive to various combinations of input parameters and different past values sequences. This makes it a potentially good candidate to detect sharp market movements. We finally mention current drawbacks and sketch how to improve them.