An Interpolation Comparative Analysis for Missing Internet Traffic Data

Abstract

In this paper, we compared interpolation techniques for fixing the missing of internet traffic data. These techniques consist of 2-D interpolation, k-nearest neighbors (KNN), and correlation. We test various types of missing data with missing probability of 0.02 and 0.98. The interpolation technique performance is shown using Normalized Mean Square Error (NMSE) parameters. Simulation results show that all interpolation methods have the ability to recover lost information. 2-D interpolation and KNN suitable for small probability of missing while correlation is suitable for large probability of missing.