CoSFuC: A Cost Sensitive Fuzzy Clustering Approach for Medical Prediction

A fuzzy clustering approach named CoSFuC was proposed in this paper for computer-aided diagnosis. Due to the predication in medical analysis was cost imbalance, and collects a large number of carefully labeled or diagnosed cases will be expensive, CoSFuC was emphasized on minimizing the misclassification cost instead of maximizing the classification accuracy, which also use the labeled and unlabeled data together to decrease the data collection burden. Experiments on eight UCI data sets (including 3 medical data sets) showed that, this method work effectively and could be used as an assistant approach for medical analysis in some circumstances.