A Comparison of Missing Data Handling Techniques

Missing data is a regular concern on data that professionals have to deal with. Efficient analysis techniques have to be followed to find interesting patterns. In this study, we are comparing 16 different imputation methods namely Linear, Index, Values, Nearest, Zero, slinear, Quadratic, Cubic, Barycentric, Krogh, Polynomial, Spline, Piecewise Polynomial, From derivatives, Pchip and Akima. These techniques are performed on real time UCI dataset and are under Missing Completely at a Random (MCAR) assumption, our result suggests the nearest, zero, quadratic and polynomial imputation methods which provides above 96% of accuracy when compared to the other techniques.