Abstract: Machine learning and data mining retort heavily on a large amount of data to build learning models and make predictions. There is a need for quality of data, thus the quality of data is ultimately important. Many of the industrial and research databases are plagued by the problem of missing values. A variety of methods have been developed with great success on dealing with missing values in data sets with uniform attributes. But in real life dataset contains heterogeneous attributes. In this paper, apart from the overview of imputation, then discussing about the proposed work i .e a new setting of handling missing data imputation (that is imputing missing data in data sets with mixed attributes and also in clustered data sets only with continuous attributes) in non-parametric mixture kernel based.