TABLE 2.

Various approaches of missing value imputation.

Approach	Advantages	Limitations	Methods	References
Global	Optimal performance when data is homogeneous	Poor performance when data is heterogeneous	BPCA	Jörnsten et al. (2005), Oba et al. (2003), Souto et al. (2015)
			SVD	Troyanskaya et al. (2001)
			ANNImpute	García-Laencina et al. (2008)
			RNNImpute	Bengio and Gingras (1995)
Local	Optimal performance when data is heterogeneous	Poor performance when data is homogeneous	KNNImpute	Dubey and Rasool (2021), McNicholas and Murphy (2010), Pan et al. (2011), Ryan et al. (2010)
			LSImpute	Bo et al. (2004)
			SVRimpute	Wang et al. (2006)
			GMCImpute	Ouyang et al. (2004)
Hybrid	Optimal performance regardless of local or global correlation	Sub-optimal performance when data is noisy and has high missing rates	LinCmb	Jörnsten et al. (2005)
			EMDI	Pan et al. (2011)
			RMI	Li et al. (2015)
			VAE, DAPL	Qiu et al. (2020), Qiu et al. (2018)
Knowledge-assisted	Optimal performance in presence of noisy data	Sub-optimal performance when data has high missing rates	iMISS	Hu et al. (2006)
			GOImpute	Tuikkala et al. (2006)
			POCSimpute	Gan et al. (2006)
			HAIimpute	Xiang et al. (2008)