چكيده به لاتين
In this dissertation, the multiperiod portfolio optimization, mean - variance is considered by downside risk measures instead of variance and taking into account the constraints that lead to the approach to real investment conditions. In the issue of optimizing the portfolio, it is decided to allocate a percentage of the total value of the portfolio to each of components with a certain return and risk. The optimization of the single-period investment portfolio, which is one of the classic issues of the field of finance, is based on three constraints: the short-term investment horizon, definite and predetermined parameters, and lack of attention to transaction costs. But in a multi-period optimization, we manage the portfolio of funds against the ongoing changes in the financial market by balancing assets, at specified intervals, to achieve maximum returns and minimize risk.
In a multi-objective optimization problem, since the objective functions often interact with each other, it is impossible to optimize all objective functions at the same time. Instead, a set of the best solutions known as an efficient frontier is achieved and allows many choices for the decision maker who chooses the appropriate solution for a particular purpose.
For this purpose, the following research has been carried out in two phases of the study: Phase I) Evaluation and selection of effective stocks that in this phase, using the data envelopment analysis, identify effective stocks from the total number of stocks in the stock market. and they are the investment candidate in the second phase of the research. Phase II) Decide on the amount of investment in each of the effective shares that passed through the first phase filter. The optimization method in this study is the method of the mean-value at risk. Therefore, the proposed model is presented as the mean model of the value-at-risk range and, after its modeling, it is solved using a non-clustered genetic algorithm (NSGA-II).
Keywords: Multi-period portfolio optimization, generalized autoregressive conditional heteroskedasticity variance models, data envelopment analysis, Non-dominated Sorting Genetic Algorithm, transaction Costs
