A New Pattern Search Method for Detecting and Forecasting Portfolio Risk Using Local Alignment Technique

Abstract: It is very important to minimize the risk in portfolio selection. For minimizing risk of portfolio at a given expected returns, it is efficient to compose portfolio with stocks which have low cross-correlation among them. In this regard, forecasting the cross-correlations among stock prices has attracted much interest among investors and financial market researchers. Most of studies investigating crosscorrelations among stock prices assume that the cross-correlations are static. But the cross-correlations in real stock markets are changing over time. Thus the dynamic clustering property and ensemble average can provide very useful methodology to calculate the time-dependent cross-correlations. In this paper, we investigated the dynamics of cross-correlations among stock prices using local alignment technique. Local alignment technique of bioinformatics can be applied to analyzing stock market and detect similar patterns in the fluctuation of stock prices during a certain time period. This information can be utilized in analyzing and forecasting the timedependent cross-correlations. This paper applied local alignment technique to the time series data of Korean stock market and found that an investor can forecast the risk of his asset portfolio and manage it optimally.