Missing data is a widely recognized problem affecting large database in data mining. The substitution of mean values for missing data is commonly suggested and used in many statistical software packages, however, mean substitution lead to large errors in correlation matrix and therefore degrading the performance of statistical modeling. The problems arises are biasness of result data base, inefficient data in missing data when anomalous data is also present. In proposed work there is proper handling of missing data values and their analysis with removal of the anomalous data.This method provides more accurate and efficient result and reduces biasness of result for filling in missing data. Theoretical analysis and experimental results shows that proposed methodology is better.