Využití stochastického programování v metodách strojového učení

Kozmík, Karel

This dissertation explores the application of machine learning methods in stochastic programming, with a primary focus on portfolio optimization. The first part investigates robust portfolio optimization with second-order stochastic dominance constraints, where uncertainty in return distributions is handled using a robust approach based on worst- case distribution modeling. The second part addresses parameter calibration in portfolio optimization models by employing machine learning techniques, specifically XGBoost, to predict optimal values that enhance out-of-sample performance. In the third part, a reinforcement learning framework is introduced for dynamic scenario tree structure selec- tion in multistage portfolio optimization, allowing adaptive representation of uncertainty. Finally, the fourth part applies multivariate probabilistic forecasting to electricity prices in the UK's day-ahead market, developing a risk-adjusted trading strategy that lever- ages price dependencies between multiple exchanges. Collectively, these contributions demonstrate the potential of integrating machine learning with stochastic programming to improve decision-making under uncertainty.

host :: přihlásit Digitální repozitář
		Hledej		Nový záznam		Nápověda		O repozitáři