An alternating least square based algorithm for predicting patient survivability
Breast cancer is the most common cancer to females worldwide. Using machine learning technology to predict breast-cancer patients' survivability has drawn a lot of research interest. However, it still faces many issues, such as missing-value imputation. As such, the main objective of this paper is to develop a novel imputation algorithm, inspired by the recommendation system. More precisely, features with missing values are regarded as items to be evaluated for recommendation.
Consequently, a matrix factorisation algorithm (Alternating Least Square, ALS) is employed to replace missing values; accordingly, four different prediction strategies based on the ALS result are further discussed. The proposed ALS-based imputation algorithm is evaluated by using a large patient dataset from the Surveillance, Epidemiology, and End Results (SEER) program. Experimental results demonstrates a significant improvement on the survivability prediction, compared to existing methods.