A Machine Learning Approach to Predict Healthcare Cost of Breast Cancer Patients
MetadataShow full item record
This paper presents a novel machine learning approach to per- form an early prediction of the healthcare cost of breast cancer patients. The learning phase of our prediction method considers the following two steps: i) in the first step, the patients are clustered taking into account the sequences of ac- tions undergoing similar clinical activities and ensuring similar healthcare costs, and ii) a Markov chain is then learned for each group to describe the action- sequences of the patients in the cluster. A two step procedure is undertaken in the prediction phase: i) first, the healthcare cost of a new patient’s treatment is estimated based on the average healthcare cost of its k−nearest neighbors in each group, and ii) finally, an aggregate measure of the healthcare cost estimated by each group is used as the final predicted cost. Experiments undertaken reveal a mean absolute percentage error as small as 6%, even when half of the clinical records of a patient is available, substantiating the early prediction capability of the proposed method. Comparative analysis substantiates the superiority of the proposed algorithm over the state-of-the-art techniques.