A portfolio optimization model based on information entropy and fuzzy time series

The traditional portfolio models assume securities' returns are normally distributed and the future distribution of returns is the same as the historical distribution. For the two stringent assumptions, this paper develops two entropy-based portfolio optimization models which are flexible and effective in measuring risk. The second motivation of this paper is to combine the fuzzy time series technique to portfolio optimization. In fact, forecasting securities' returns distribution is an important issue for portfolio. And among the many forecasting methods, the fuzzy time series technique is more suitable to deal with the fuzzy data in financial data. The empirical results on the historical data of the Stock Exchange in Chinese financial market show effectiveness of the proposed models. Both the entropy based models outperform the traditional ones and the fuzzy time series forecasting model also helps to further improve the actual performance.