Ohodnocení vlivu částečné informace u stochastických optimalizačních modelů s penalizací

Abstract

Stochastic optimization models with penalties are essential for businesses when deciding on the optimal production quantity to achieve the highest possible profit. Current models utilize knowledge about possible scenarios and the likelihood of their occurrence. This work focuses on extending these models with partial information obtained through neural networks, expert evaluations, or access to databases. The aim of this study is to explore how partial information about the probability of scenario occurrence in individual periods can improve stochastic optimization models with penalties and what the impacts on business decision-making are. Theoretical foundations of stochastic optimization, solution methods of stochastic optiization, sensitivity analyses, and the application of these approaches to two textbook examples, the Farmer's Problem and the News Vendor Problem, are examined in this study. The simple sensitivity analysis method is used for the analysis. The analysis of the results revealed that the expected profit increases faster than linearly with increasing reliability of partial information. It was also found that the value of partial information varies depending on the probability of scenarios. This study demonstrates the potential advantages of using the value of partial information when deciding to acquire partial information and provides businesses with guidance on incorporating partial information into their decision-making models and evaluating partial information. Further research focused on incorporating the changing price of partial information and the possibility of altering the probabilities of individual scenarios is recommended.