The agribusiness has great importance for the Brazilian economy, representing a significative share of GDP and total exports of the country. Like other production processes inserted in an environment of uncertainty, the agricultural activity needs instruments that minimize the risk, especially, the risk of price and assist in decision-making process of agents participating in agribusiness. In this context, the future markets constitute themselves as financial alternatives in risk management through the hedge operations. However, the efficiency of these operations depends on the application of appropriate methodologies that lead to more precise knowledge about future prices. Thus, the main objective of this work is to evaluate the application of jumps diffusion models, so successful for the structure of interest rate, in the case of commodities; focusing on the achievement of forecast. The empirical analysis will be carried out from the historical range of arabic coffee agricultural commodity´s prices traded on BM&F. The methodology used is based on the paper from Sanjiv Das (1998). In this thesis it will be estimated a class of stochastic models described in the literature, such as the mean-reversion process, the geometric Brownian motion and its variants with jumps. ARCH and GARCH effects will also be considered in the modeling. The process of estimation will be developed by traditional methods as well as by Genetic Algorithms. Facing the problem and the shortage of modern studies directed to the approach of the agricultural commodities in the country, the topic proposed is a motivation for scientific and technological research.