Selection of number of factors and neighbours in global-local iterative least squares data imputation (INI) algorithm for microarrays gene expression

Due to several reasons, the DNA microarrays data collected may have a number of missing. This incompleteness data problem may arise the difficulties to apply commonly used clustering techniques such as K-means, Hierarchical clustering, Self-Organizing Map (SOM) and Principal Component Analysis (PCA). In this report, the experimental analysis of the selection of number of factors and neighbour used in global-local iterative least squares imputation algorithm, which is called as INI imputation, will be studied in the context of DNA microarrays gene expressions. The experiments have been carried out on real data set that was obtained from experimental study in identification of dif- fuse large B-cell lymphoma. The results show that for all, except in case the number of neighbour genes used less than 5, INI outperforms KNNimpute and LLS.