Hybrid approach, Missing values, K-nearest Neighbour, Singular ValueDecomposition.

Missing data arises in almost all serious statistical analyses. Statistical analyses havea variety of methods to handle missing data, including some relatively simpleapproaches that can often yield reasonable results such as the random imputationapproach. The missing data imputation process must be modeled in order to performimputations correctly. Using datasets in empirical applications is very common toperform some tasks; however, missing values in datasets should be extracted fromthe datasets or should be estimated before they are used for processing to producecorrect association rules or clustering in the preprocessing stage of data mining andprocessing. In this thesis, a hybrid approach is used that combines K-NearestNeighbor (KNN) with Singular Value Decomposition (SVD) algorithm to improvethe data imputation and produce data with high correlation with original missingvalues. The test results of the proposed hybrid method are compared with the resultsof several alternative methods for different rate of missing values and the results ofthe proposed method yields better performance than the others. The results are alsocompared with the reported results in the literature to give an idea about itsperformance.