关键词: K-means聚类算法, 广义熵, CVaR模型, 投资组合

Abstract:

The present work constructs the CVaR linear programming model of portfolio with the constraint of generalized entropy. We generate scenarios and probabilities of each asset yield in the portfolio using the K-means clustering method. Then we substitute them into the model. Finally we get the optimal investment weights for various assets. The feasibility of this model is certificated by testing a portfolio which contains eight selected stocks in Shenzhen stock market. Compared with MV model, this model not only incorporates more decentralized investment principle, but also has better performance in the future yields. This model has strong practicability.

Key words: K-means clustering method, generalized entropy, CVaR model, portfolio