A Fitted Fuzzy-rough Method for Missing Data Imputation

Abstract

Missing data imputation is a fundamental task for reducing uncertainty and vagueness in medical dataset. Fuzzyrough set has taken very important role to accurate representation original information. This paper proposes Fitted fuzzy-rough imputation algorithms called Fitted FRNNI and Fitted VQNNI by introducing weight coefficients to balance fuzzy similarly relations among training and testing instances. Meanwhile, modification fuzzy decisions of nearest neighbors based on lower/upper approximations are studied. Performance analysis is conducted including classification accuracy analysis, the impact of k parameter and weight coefficient of \alpha and \beta to evaluate the proposed Fitted FRNNI and VQNNI algorithms. Experimental results on 13 benchmark datasets show that the proposed algorithms outperform current leading algorithms. 2019 IEEE.