TABLE 3.

List of data transformation and feature scaling techniques prior to dimensionality reduction.

Type	Advantages	Limitation	Technique	Reference
Normalization	Identifies and removes systematic variability. Increases the learning speed.	Less effective if high number of outliers exist in the data.	Quantile	Larsen et al. (2014)
				Smyth and Speed (2003)
				Schmidt et al. (2004)
			Loess	Franks et al. (2018)
				Karthik and Sudha (2021)
				Larsen et al. (2014)
				Huang et al. (2018)
				Bolstad et al. (2003)
				Doran et al. (2007)
Data transformation	Reduces the variance and reduces the skewness of the distribution of data points.	Data do not always approximate the log-normal distribution.	Log transformation	Pirooznia et al. (2008)
				Pan et al. (2002)
				Doran et al. (2007)
Standardization	Ensures feature distributions have mean = 0. Applicable to datasets with many outliers.	Less effective when data distribution is not Gaussian, or the standard deviation is very small.	z-score	Peterson and Coleman (2008)
				Cheadle et al. (2003)
				De Guia et al. (2019)
				Chandrasekhar et al. (2011)
				Pan et al. (2002)