In this paper, we describe a prediction method applied for the evaluation of the troposphere ozone concentrations in the Romanian city of Bucharest; in this area, the pollution level is very high principally due to the traffic and multiple local sources. Using only meteorological data and some precursors of the ozone we estimate the ozone concentrations for a horizon of 24 hours. The minimization algorithm used here is the Levenberg-Marquardt algorithm in the framework of the Scilab program. Sensitivity tests performed show the robustness of the results and a good applicability of this model for a short-term prediction of the ozone.