Breast Cancer Prediction Using Neuro-Fuzzy Systems

Cancer is one of the most dangerous diseases in the world. The scientists are in pursue of finding better methods of detecting the various type of cancerous cell formations in the tissues. The purpose of this work is to develop a more accurate prediction model to identify breast cancer. In this work, Genetic algorithm (GA) based trained recurrent fuzzy neural network (RFNN) and adaptive neuro-fuzzy inference system (ANFIS) are used on the dataset provided by the UCI Machine Learning Repository. In this data set there are 9 quantitative attributes and a label that clinical features are observed or measured for 116 participants. The dataset separated into two sub-sets; one for training (81 instances) and one for testing (35 instances). For 8 different combinations of variables 8 different GA based trained RFNN and 8 different ANFIS were designed. The sensitivity, specificity, precision, F-score, probability of the misclassification error (PME) and accuracy of the training set, testing set and overall performances of the models were analyzed. The RFNN with 9 variables gave the highest overall accuracy (88.79%). The overall results showed that the GA based trained RFNN outperformed both ANFIS and other previous works that used the same dataset.