Prediction of radical hysterectomy complications for cervical cancer using computational intelligence methods

Abstract

In this work, eleven classifiers were tested in the prediction of intra- and post-operative complications in women with cervical cancer. For the real data set the normalization of the input variables was applied, the feature selection was performed and the original data set was binarized. The simulation showed the best model satisfying the quality criteria such as: the average value and the standard deviation of the error, the area under ROC curve, sensitivity and specificity. The results can be useful in clinical practice. © 2012 Springer-Verlag Berlin Heidelberg.