Author: Valentin Burca
In the light of the current budget constraints, the investors face a challenge when building their stock portfolios that should lead to a minimized risk for an expected level of return. Mathematical tools have become essential for portfolio theory formulation in the last decades. In this article, our main objective is to illustrate the utility of some data mining tools and techniques, with a focus on principal components analysis and cluster analysis. The case study reveals a comparative empirical results analysis of the classical Markowitz portfolio optimization model and a combined data mining techniques model. The results show how useful data mining techniques can be for the finance area, with positive implications for investors’ strategy design and implementation. Our study reveals that stocks selection requires the use of modern techniques that take in account the multidimensional perspective of investment decision. Henceforth, we propose that a debate should be launched concerning the design of stock markets design, which generally focus on simple design oriented to the stocks liquidity. In order to help investors, those indices should combine multiple dimensions of stocks definition, as the return, the risk and the liquidity of a stock are at least of the same importance from an investment decision perspective.