Time Series Segmentation and Statistical Characterisation of the Spanish Stock Market Ibex-35 Index

Abstract

The discovery of characteristic time series patterns is of fundamental importance in financial applications. Repetitive structures and common type of segments can provide very useful information of patterns in financial time series. In this paper, we introduce a time series segmentation and characterisation methodology combining a maximal likelihood optimisation procedure and a clustering technique to automatically segment common patterns from financial time series and address the problem of stock market prices trends. To do so, the obtained segments are transformed into a five-dimensional space composed of five typical statistical measures in order to group them according to their statistical properties. The experimental results show that it is possible to exploit the behaviour of the stock market Ibex-35 Spanish index (closing prices) to detect homogeneous segments of the time series.