Abstract: Breast cancer is the most common cancer that affects women in the world, and predicting breast cancer viability is a complex and challenging work. Most previous efforts focused on statistical or supervised approaches to predict the survival prospects of patients. However, both complex feature correlation and missing feature values may increase the difficulty of survivability prediction. In this paper, we propose a novel hybrid model to ameliorate patient survivability analysis. Firstly, we utilize principal component analysis (PCA) and K-means to create clusters with similar characteristics based on the different stages of breast cancer spread. Then, these clusters are exploited to train the classification model for patient survivability prediction. Experimental results show that, compared with the original historical data, the accuracy of survivability prediction for specific models is further improved by using identified patient cohorts.

Key words: breast cancer survivability prediction, PCA, K-means, hybrid model