Combination of KNN-Based Feature Selection and KNNBased Missing-Value Imputation of Microarray Data

Abstract

Microarrays are useful biological resource to study living forms at the molecule level. Microarrays usually have only few samples but high dimensionality with many missing values. The consequent downstream analysis becomes less efficiency. This paper proposes a methodology to impute missing values in microarray data. The proposed methodology is a combination of KNN-Based Feature Selection and KNN-based imputation (KNNFS Impute). The KNNFS Impute comprises of two main ideas: feature selection and estimation of new values. A comparative study of the proposed method with traditional KNN and Row average methods has been presented for the estimation of the missing values on three microarray data sets: Lung Tumor, Colon Cancer, and ALL-AML Leukemia dataset. The best estimation results are measured by the minimum Normalized Root Mean Squared Error (NRMSE). The results show that the proposed method has powerful estimation ability on the three data sets with smaller NRMSE than the compared methods. The 3rd Intetnational Conference on Innovative Computing Information and Control (ICICIC'08) 978-0-7695-3161-8/08 $25.00 © 2008 IEEE The 3rd Intetnational Conference on Innovative Computing Information and Control (ICICIC'08) 978-0-7695-3161-8/08 $25.00

Missing data imputation with fuzzy feature selection for diabetes dataset

“…As a consequence, the performance of KNN degrades, especially when the missing rate increases [9]. The paper proposes feature selection before imputation which is the modified KNN called KNN-based feature selection (KNN-FS) [10]. It is then found out that, by implementing feature selection before imputation, the proposed method performed better than traditional KNN in terms of NRMSE when applied to three microarray datasets: Lung Tumor, Colon Cancer, and ALL-AML Leukemia dataset.…”

K-Means Clustering on 3^rd order polynomial based normalization of Acute Myeloid Leukemia (AML) and Acute Lymphocyte Leukemia (ALL)

“…K Nearest Neighbor is a very simple algorithm to impute missing values in a correlated Microarray data. It matches the similar expressions levels of a gene in its neighbor hood [13][14][15] and assumes that the expression levels missing values are very close to each other. This procedure is given below in steps;…”

Learning k for kNN Classification

“…Meesad and Hengpraprohm proposed an imputation method combing the kNN-based feature selection with kNN-based imputation. Differing from the conventional kNN method, their method first conducts feature selection, and then estimates missing values [Meesad and Hengpraprohm 2008]. García-Laencina et al proposed to employ mutual information to design a featureweighted distance metric for conducting kNN [García-Laencina et al 2009].…”