Distribution-Based Entropy Weighting Clustering of Skewed and Heavy Tailed Time Series

The goal of clustering is to identify common structures in a data set by forming groups homogeneous objects. observed characteristics many economic time series motivated the development classes distributions that can accommodate properties, such as heavy tails and skewness. Thanks its flexibility, skewed exponential power distribution (also called generalized error distribution) ensures unified general framework for possibly tailed series. This paper develops procedure model-based type, assuming are generated same underlying probability but with different parameters. Moreover, we propose optimally combine estimated parameters form clusters an entropy weighing k-means approach. usefulness proposal shown means application financial series, demonstrating also how obtained be used portfolio stocks.