Breast Cancer Classification Using Fuzzy Elman Recurrent Neural Network

Dhoriva Urwatul Wutsqa and Anisa Nurjanah

In this study, we propose a new approach, namely fuzzy Elman recurrent neural network (FERNN), for breast cancer classification. The structure of FERNN is constructed by integrating the fuzzy concept and Elman recurrent neural network (ERNN), a type of neural network model which has a partial feedback connection. The fuzzification process, which uses membership function, is integrated into the ERNN model by converting the crisp inputs into fuzzy inputs. The breast classification in this model uses the parameters of Wisconsin Breast Cancer (WBC) and Wisconsin Diagnostic Breast Cancer (WDBC) database. The FERNN is then evaluated in a cross-validation setting in terms of three criteria including accuracy, sensitivity, and specificity. The experiment results demonstrated that the FERNN yields excellent performance. The FERNN model on WBC data gives the classification value of 100% for all criteria, which are accuracy, sensitivity, and specificity, for both training and testing data. Furthermore, the model on WDBC data gives the classification accuracy, sensitivity, and specificity of 97.58%, 94.70%, and 99.29% for training data, respectively, and all 100% for testing data. It also shows the superiority of FERNN to the most existing methods.

Volume 11 | 11-Special Issue

Pages: 946-953

DOI: 10.5373/JARDCS/V11SP11/20193119