Prediction of Breast cancer survivability using semi supervised machine learning framework

Artificial intelligence (AI) and machine learning (ML) algorithms are gaining acceptance in the field of health care analytics due to their effectiveness in terms of precision and prediction accuracy. A semi-supervised machine learning framework is proposed in this paper that is based on cluster-and-then-classify approach. The sole purpose of this approach is to enhance the accuracy of breast cancer survivability prediction. Stochastic optimization algorithm has been used for optimal feature selection. Famous standalone clustering algorithms have been implemented in the framework followed by classification step. The clustered output acts as an expert view and can be used in the classification process thereby optimizing the model’s performance. The experiments show that blending of clustering results with the optimal features better train most of the standalone supervised machine learning classifiers. Based on the analysis, it is endorsed that compared to the existing cutting-edge ML techniques, the proposed framework using cluster-and-then-classify approach has resulted better performance. The logarithmic – loss has been recorded as low as in the range of [0.9.99E-16 – 0.85]. The area under the ROC curve (ROC-AUC) has been increased up to the range of [0.99 - 1.0]. The prediction accuracy has been enhanced up to the range of [96% -100%].