An imputation-based method for fuzzy clustering of incomplete data

Several data mining processes have to deal with incomplete data which hampers the quality of output of analysis techniques. A few imputation-based techniques have provided good methods to deal with missing data. Yet all of them are of impute-and-analyze nature. This paper proposes an iterative technique to impute data and fine tune the estimates. It is based on the idea that information from complete objects should be garnered to estimate missing values such that purpose of analysis is supported. Based on statistical information granules, the proposed method is a fusion of imputation technique and fuzzy clustering.