DBSCANI: Noise-Resistant Method for Missing Value Imputation

Abstract

The quality of data is an important task in the data mining. The validity of mining algorithms is reduced if data is not of good quality. The quality of data can be assessed in terms of missing values (MV) as well as noise present in the data set. Various imputation techniques have been studied in MV study, but little attention has been given on noise in earlier work. Moreover, to the best of knowledge, no one has used density-based spatial clustering of applications with noise (DBSCAN) clustering for MV imputation. This paper proposes a novel technique density-based imputation (DBSCANI) built on density-based clustering to deal with incomplete values in the presence of noise. Density-based clustering algorithm proposed by Kriegal groups the objects according to their density in spatial data bases. The high-density regions are known as clusters, and the low-density regions refer to the noise objects in the data set. A lot of experiments have been performed on the Iris data set from life science domain and Jain’s (2D) data set from shape data sets. The performance of the proposed method is evaluated using root mean square error (RMSE) as well as it is compared with existing K-means imputation (KMI). Results show that our method is more noise resistant than KMI on data sets used under study.

A systematic review of machine learning-based missing value imputation techniques

“…The details of this categorization are presented in Table 1. Simple and easily implemented (Lu et al, 2014) Best fit missing value imputation (BFMVI) Significantly less complex but accurate imputation method for Internet of Things (IoT) data sets (Agbo et al, 2020) Cluster-directed framework for neighbor-based imputation Enhance the accuracy by imputing missing values in microarray gene data (Keerin et al, 2016) C-means Identify the similar records in the complete data set and find the best value to replace the missing value in the data set (Samat and Salleh, 2017) Handles missingness in distributed data sets (Sefidian and Daneshpour, 2019) Imputation utilizes feature selection to select only highly relevant features (Himmelspach and Conrad, 2010) Correlated cluster 100% accuracy during experimentation (Myneni et al, 2017) Density-based clustering Predict missing values correctly in noisy environment (Purwar and Singh, 2016) Steady imputation performance (Razavi-Far and Saif, 2016)…”

Transcriptional signatures of the BCL2 family for individualized acute myeloid leukaemia treatment

“…In brief, for the gene selection step, we identified genes closely related to the BCL2 family from a curated gene-set database [ 36 – 62 ] (Additional file 1 ). In each AML dataset, subsequently, a backward selection was conducted to remove high-noise genes with a postulation that they may not be helpful in the imputation of the profiles of the BCL2 family genes (BCL2, MCL1, BFL1, BCLXL, and BCLW) [ 63 , 64 ] (Additional file 2 : Fig. S1).…”