This paper presents comparison of genetic algorithms (GAs) and evolutionary programming (EP) to estimate parameters of a linear auto regressive model with external input (ARX) and an auto regressive moving average model with external input structures (ARMAX) that predict the behavior of air temperature within a greenhouse. Data groups were used to estimate and validate models and these data groups were 20:80, 33.33:66.67, 50:50, 66.67:33.33 and 80:20%. The objective was to determine which evolutionary algorithm generates parameter values that give the best prediction of the environment in a greenhouse located in the central region of Mexico. Simulation and analysis of the ARX and ARMAX model’s performance show that these models under-estimate measurements. Furthermore, the estimations of the inside temperature have a better fit when the parameter identification of an ARX structure is calculated by means of GAs, so that, there is a better fit of the simulated data to measured data when the 20% of the data are used to estimate and 80% of the data are used to validate the model. Key words: Auto regressive moving average model with external input structures (ARMAX) model, auto regressive model with external input (ARX) model, genetic algorithms, evolutionary programming, parameter identification.