In the present work, the CRISP-DM methodology was proposed to develop a set of machine learning models applied to evaluate cervical cancer risk suffer. For this research, a sample of 858 patients was taken, who were asked a series of questions regarding this pathology. The database has an unbalanced dependent variable, since this is a health study, the balancing technique will not be used to identify the variables that will enter the model, the Boruta library was used for variable selection. For the development, five algorithms will be used: Support Vector Machine (SVM), decision trees using the CHAID and CART algorithms, logistic regression and "asymmetric link" models. The models proposed in this work were refined by means of the Auc, Gini, Log loss and KS (Kolmogorov-Smirnov) indicators, as a result using the proposed models, AUC values of 98% were obtained.