Performance comparison of State-of-the-art Missing Value Imputation Algorithms on Some Bench mark Datasets

The presence of missing values influences the selection of appropriate set of attributes that render degradation in classification accuracies of the classifiers. Missing values are a common problem in almost all real world data sets [1] used in knowledge discovery and data mining(KDD) applications. Specifically they are more frequent in clinical databases [2, 3, 4] and temporal climate databases [5, 6]. Their presence would greatly affect the performance of classifiers [7]. The missing values in the databases may arise due various reasons such as the value being lost (erased or deleted) or not recorded, incorrect measurements, equipment errors, or possibly due to an expert not attaching any importance to a particular procedure. The incomplete data can be identified by looking for null values in the data set. However, this is not always true, since missing values can appear in the form of outliers or even wrong data (i.e.