Nonlinearity linkage detection for financial time series analysis

Standard detection algorithms for nonlinearity linkage fail when applied to typical problems in the analysis of financial time-series data. We explain how this failure arises when standard algorithms are applied na¨ıvely, how linkage detection needs to be applied directly to the observed data samples, and how this raises problems that are not addressed by current techniques. We extend the existing DSMDGA linkage detection algorithm and present a new system that can determine the required nonlinearity linkage in observed data samples for financial time series. The new system has been evaluated on synthetic datasets and experimental results are provided. The sensitivity of the system to changes in both the problem and the algorithm parameters has also been explored and we discuss the results. We present evidence of the success of the new system and identify areas for further work. Categories and Subject Descriptors I.2.M [Artificial Intelligence]: Miscellaneous General Terms Algori...

• Type: Conference
• Year: 2007
• Where: GECCO
• Authors: Theodore Chiotis, Christopher D. Clack
• Keywords: GECCO 2007, Linkage Detection, Nonlinearity Linkage, Observed Data Samples