Over the past four thousand years, numerous techniques have been developed and used to address problems in Finance. These techniques include simple arithmetic calculations and probabilistic methods as well as intelligent systems techniques such as neural networks, genetic algorithms, multi-agent systems, and support vector machines. The techniques have been developed to accurately and quickly collect, validate, analyze, and integrate data that change dynamically. The particular problem that we address in this dissertation is the construction of efficient algorithms for the problem of an optimal portfolio selection, that is, algorithms that would accurately and in real time determine the best distribution of wealth among several investment assets to achieve a specific goal. The main goal of making investments is to gain as much wealth as possible, therefore the goal is to maximize the return. However, the problem we face is that the investors do not know in advance what the return of each asset will be, but rather there are some educated predictions about the returns from each asset. These return predictions might turn out to be correct, but on the other hand, they might be completely wrong. Thus, there is a risk associated with each predicted return, and therefore with each asset. One of the goals of an investor is the minimization of risk. However, usually, a higher return is associated with a higher risk. Thus, it is impossible to maximize the return and minimize the risk independently. Moreover, additional characteristics, such as time to maturity, preferred portfolio structure, and reputations of companies that sell investments asset, among others, are also considered before locking money into an asset or a portfolio. This problem leads to the necessity of developing intelligent system techniques to find the best trade off between the return and the risk based on the preferences of an investor. The techniques that are used to solve the problem of optimal portfolio selection all have their drawbacks. To solve these problems, we propose a new approach driven by utility-based multi-criteria decision making setting, which utilizes fuzzy measures and integration over intervals. [PUBLICATION ABSTRACT]
