Automatic method for the prediction of the commercial appraisal of a property in Bogota city

In this paper was done design, develop and implement of an automatic method by machine learning techniques with which it can predict values of real state housing in Bogota city, through extracting data from sell real state publication in web pages. With the tool for web scrapping “Dexi” was extracted information from de web page “Finca Raiz”. Then was made a cleaning process of data with help of Python and Excel. After, was implemented the techniques of: Decision tree, lineal regression, Random Forest and deep neuronal network. Lastly, was calculated performance measures of coefficient of determination (R2), Root mean square error (RMSE) and coefficient of variation (CV) respect to variance, for select the best method. As a result, is obtained an organized dataset with the information extracted from web page “Finca Raiz”. However, by the conditions of heterogeneity of these data, Dataset was segmented by social stratum and algorithms was implemented by each one. For random forest was obtained a mean of coefficient of determination of R91% and coefficient of variation of 17%, being this for present investigation the best model of machine learning when obtaining performance measures better than the rest models.