Combining K-Means and Gaussian Mixture Model for better accuracy in prediction of Ductal Carcinoma in Situ (DCIS)- Breast Cancer

The most frequent malignancy in women now a days is breast cancer particularly in India. To predict DCIS( Ductal Carcinoma in Situ) various trained model of supervised learning algorithms like k-Nearest neighbor, Decision Tree classifiers, Random forest classifiers are used by various researchers. In unsupervised learning model K-means algorithm is also used to classify the clinical characteristics, mammographic findings, and images features. In this paper K-means (unsupervised learning) idea has been extended to combine the Gaussian Mixture model (GMM) for better accuracy. The EM (Expectation Maximization) based GMM is combined with k-means for flexibility in cluster shape by density estimator to overcome the problems of k-means algorithm. The simulated results show the competence of the proposed model to predict for the early diagnosis of DCIS (the early stage of the breast cancer). Scikit-learn, Pandas, Matplotlib and Numpy modules of Python have been used for breast cancer datasets, simulation and results.