This research presents the analysis and comparison of prediction models for the market price of real states by means of applying machine learning schemes based on artificial neural networks (ANN). Three prediction models are developed, each considering a particular set of attributes and a different ANN architecture. The research is carried out following a 4-stage methodology: analysis, design, construction, and validation. To construct the prediction models, 228 real estates are considered. For each property, 35 attributes were documented. The dataset was split into 2 files. The first one containing 80% of the data for training and testing purposes, while the remaining 20% is left for validation. To reduce the uncertainty, a cross validation strategy was applied. All prediction models are finally compared by means of the error measures MAPE, MAE, and RMSE. In all cases, prediction results present a MAPE between 14% and 16%. In conclusion, the research revealed promising results when machine learning schemes are used to predict real estates’ market prices. pricing model.