Risk Decision Model Of Multistage Project Portfolio

Generally, the venture capital firms or venture capitalists often use portfolio and multi-stage investment strategies for the sake of avoiding being "held up" by venture business. In investment activities, the invest costs of projects are uncertain, which were predicted from the historical data s or gathered from the experts, and they usually differs a lot from the realized projects' costs. This is why it is hard to choose the true optimal portfolio that minimizes the realized ex post costs. In this dissertation, we start from considering the uncertainty of the projects' value or invest cost, based on the latest theoretical research results at home and abroad integrate using of venture capital theory, stochastic optimization theory, Bayes method and statistic simulation technology, combine portfolio investment with multi-stage investment strategies and study the multi-stage project portfolio problem under the uncertain environment. Specifically, the main contents of this paper are as follows:Therefore, in chapter two, we developed a Bayesian model to support portfolio cost minimization in the presence of uncertain cost estimates, for minimizing the portfolio‘s total cost. We firstly apply Bayes method to estimate the cost of investment, use of the revised estimate for portfolio selection,and eliminate the expected positive gap between the true and estimated portfolio cost, i.e., post-decision disappointment. In addition, Bayesian analysis can also be used to study the expected value of obtaining additional cost estimates for the projects prior to actually acquiring these estimates. Analytic results were derived for expected value of additional information in the case where projects' costs and cost estimates were log-normally distributed. The results show that the decision maker will have less cost if he utilizes Bayes method to estimate the projects' cost and uses of the revised estimate for portfolio selection. What's more, Bayes method can improve the accuracy of project estimates and reduce the decision maker's post-decision disappointment. Compared with the portfolio's cost which based on the revised estimate using Bayes method, it is more closely to the portfolio's cost of ture cost when making decisions based on the expected value of additional information.Secondly, based on the classical Markowitz mean- variance model, we considered a multi-stage risk investment portfolio problem under uncertainty in chapter three and set a multi-stage stochastic programming problem for choosing and managing portfolio. Hence, we use a discrete scenario tree to describe the change of alternative project value in an investment plan period. Whereafter, considered the cost of projects and the cash generated in every period, as well as followed the principle of minimizing the risk of the portfolio beyond a certain value, we established a multi-stage risk investment portfolio optimization model based on semivariance risk measure with limited resources. Finally, we analyzed a numerical example verifying that the model established is reasonable and effective. If a project did not have a good performance, the model allows decision maker to adjust the project decisions by abandoning it. By this way, the risk of portfolio can be reduced partly and the decision maker may get more profit.