The optimal portfolio with modified covariance matrix using clustering method

The Markowitz optimal portfolio was introduced and the reason why the result was unstable was analyzed.Based on this analysis, a new method was presented: Using the clustering method to modify the sample covariance matrix to get a better investment option. To prove the new methods reasonableness, real data from the Chinese stock market were used to simulate “real investment”. It was found that the portfolio obtained from this method was better in both mean return and stability than the traditional method， which can be further verified by using risk prediction.