Abstract
Clustering is an unsupervised approach for unlabeled data classification with less prior information. Despite the usefulness of data clustering, few of clustering algorithms are applicable to deal with incomplete datasets, which are datasets with missing values. In addition, this type of datasets takes the majority in real-world applications. Therefore, a new algorithm combining fuzzy c-medoids clustering and partial distance strategy is developed to overcome the problem. A physical examination case study demonstrates the advantage of the proposed algorithm. The experimental results show that data clustering outperforms the traditional dichotomy identification in revealing potential patients and can be employed in preventive medicine.