|
In all investing ways, mutual fund is one of investing tools by the most people using because of its low entry threshold. The mostly existing papers use performance of mutual funds to explore the investment allocation and selection, and the performance of mutual funds come from its net rate of return. And therefore the forecast for the net and influencing factors has become the purpose of this study. In this study, according to genetic programming's solving the problem and plannig solution of flexible capacity, we apply it to explore and model the inputting relevant variables .And hope to develop a more comprehensive and stable NAV forecast. According to the result of this study, we can sum up in the following conclusions 1. During the trainning period, fitness function using MSPE is better than MAPE, but the NAV forecast is MAPE more stable than MSPE. 2. The variables for constant、one month ROI、three months ROI、six months ROI and buying turnover rates are often used in the process of evolution. It means these five variables have influence for computing NAV. 3. If we input a single variable to evolve, it can not effectively predict or increase correct rate of the fund's change. 4. The ADF with fixed formula for modeling the NAV prediction does not improve learning ability of algorithm. Recommanding the follow-up studies can define the function inside the ADF that re-define the related algorithms or rules to conduct a study.
|