Detecting clusters and their dynamics in the Forex Market
Tutor / director / evaluatorArratia Quesada, Argimiro Alejandro
Document typeBachelor thesis
Rights accessOpen Access
This project studies and implements the clustering methods introduced by Fenn et al. to detect correlations in the foreign exchange market. To deal with the potentially non linear nature of currency time series dependance, we propose two alternative similarity metrics to use instead of the Pearson linear correlation. We observe how each of them responds over several years of currency exchange data and find significant differences in the resulting clusters.