Design and analysis of an iterative algorithm for incomplete data estimation

Abstract

We present in this paper a new technique for estimating (i.e., predicting) missing data in a data array, based on the nearest neighbours approach. This technique has the advantage of needing no matrix diagonalization and no assumptions on the initial data and may be easily extended to ternary arrays. The practical efficiency of our study is illustrated through the analysis of a (60 × 10) array of magmatic data

Keywords:

• Data estimation
• missing data
• nearest neighbours
• reconstitution
• prediction