Stochastich optimization: Option in farms?

Abstract

Abstract The results of a simple model of agricultural linear programming (PL) in Gral. Villegas were explored in a first step. The optimal plan was analyzed using Monte Carlo simulation (SMC). The random variables considered were yields and prices at harvest time. Next, stochastic optimization was used, which works thanks to the development of genetic algorithms. It makes it possible to arrive at an optimum using random variables in the same procedure. It was supposed to achieve a “synergy” effect between the (PL) and (SMC) tools. The exploration allowed to visualize the robust activities, those that are repeated invariably, although with different dimensions in the solutions. However, the most valuable results of LP (activity substitution cost, resource shadow price, and validity limits) are lost when working simultaneously with SMC. You can only count on the best solution and from it obtain the minimum, average and maximum margin. Nor are the probability distributions (mass and cumulative) of the SMC obtained. The importance of the marginal values ​​with which LP works and the correct interpretation of the different outputs is not an easy task: it requires considerable training. Simple and understandable SMC outputs of possible plans, associated with probabilities of occurrence, are also not available when working with stochastic optimization. Both features are not indicated by software vendors. It is suggested to continue the exploration to elucidate if the option allows help for decision makers.