Liovkin V. Methods and tools for investment decision-making under uncertainty

The thesis is devoted to the development of methods and tools for investment decision-making under uncertainty for increasing the efficiency of investment management process based on management characteristics prognostication improvement. The object of the research is investment management decision-making process under uncertainty, the subject of the research is methods and tools of investment decision-making under uncertainty. The integrated investment decision-making problem model, which is based on the finite horizon optimal stochastic control and enables to distribute capital between real investment and security portfolio, is stated. The problem solution method, which integrates real and portfolio investment result prognostication in the control process, includes supporting decisions mechanisms in distribution and monitoring stage, is proposed. The method of real investment unsuccessful risk prognostication, which uses ensembles of neural networks with additional diagnostic criteria (planned costs and duration of real investment process) for prognostication, is developed. Data clustering and using of a separate set of neural networks of different architectures for real investment success prognostication makes specification of neural networks in the ensemble. This method enables to classify real investments by its success level, increasing accuracy of the results, and provides a mechanism for real investment unsuccess risk evaluation. The method of prognostication of actual real investment results deviation from planned real investment results, which enables to improve the accuracy of prognostication results and to improve the process automation level, is developed. Prognostication of actual real investment costs and duration deviation from planned results is made based on data clustering using self-organizing maps and specialized cascade neural networks with feedforward signal and backpropagation error using neural-evolutionary approach for network configuration. The modified security portfolio management decision-making method based on D-scores of Russman, which allows to manage investments more efficiently on unstable stock exchange, is proposed. Investment management decision support system is developed. Application efficiency of the proposed methods and tools is confirmed by the implementation results.