A Hybrid Model for Breast Cancer Diagnosis Based on Expection-Maximization and Artificial Neural Network: EM+ANN
The correct diagnosis of breast cancer is an important subject in the medical field. Different data-mining techniques have improved a great deal to help medical experts in diagnosing diseases. This paper presents an expert medical diagnosis system for predicting breast cancer based on an EM cluster algorithm and artificial neural network (ANN). With the EM clustering method, dataset is divided to homogeneous subclusters to ensure more homogeneous training and test datasets for ANN. The proposed model EM-NN achieves more efficient training, while maintaining ANN performance. To test the proposed hybrid model, we used the Wisconsin Breast Cancer Database (WBCD) from UCI machine learning repository. The correct classification rate of EM-NN is 98.54%. This result demonstrated that to enhance the classification accuracy, the training process of the model is necessary to use homogeneous data in a learning-based classification algorithm. The proposed model can be used to obtain correct results for other diseases.
© 2011 Karaelmas Fen ve Mühendislik Dergisi