Cluster financial time series for portfolio

Stocks are a common kind of financial time series. In this paper we present a new similarity measure for time series clustering, and then select a set of stocks to create efficient portfolio, which is of crucial importance in the process of creating efficient portfolio. We largely reduce the efficient times of portfolio using clustering-based selection, and only select a subset of stocks from different groups to create efficient portfolio each time, then it is easy to get the portfolio with the lowest risk at a given level of return. A set of 100 stocks were utilized for experiments, and compared with other selection methods, the results show that our method could largely reduce the efficient times of portfolio. Group-ward hierarchical cluster was used to cluster stocks.