Model Predictive Control for Stock Postfolio Selection

The motivation behind the thesis lies in some interesting results in the article 'Portfolio Optimization Applications of Stochastic Receding Horizon Control' written by James Primbs, regarding applying control theory to problems in the financial market. The purpose will be to implement a method called Model Predictive Control, MPC, for selecting optimal portfolio weights in two portfolios with different objectives, a risk adjusted portfolio and an index tracking portfolio. The basic idea behind choosing this method when optimizing portfolios in the financial market is its ability to use feedback and in the same time includes the future when optimizing at each time step. The method also handles multivariable problems and constraints and is therefore suitable for the complex portfolio weight optimization. To see how the method works, experiments with index tracking is performed with real data provided by Danske Bank. This shows promising results for small portfolios but for larger portfolios the setup is needed to be developed further. A new approach for this adjustment is presented. Some further research will be needed to be able to handle problems in reality in a good way and this will be discussed. The thesis holds all needed definitions regarding the modeling technique and terms used