TABLE 4.

List of different feature selection and feature extraction techniques.

Approach	Advantages	Limitation	Feature Selection Techniques	Reference
Filter	Datasets are easily scalable. Perform simple and fast computation. Independent of the prediction- outcome. Only one-time feature selection.	Ignores the interface with the classifier. Every feature is separately considered. Ignores feature dependencies. Poor classification performance compared to other feature selection techniques.	t-statistics (t-test)	Pan et al. (2002), Önskog et al. (2011)
			Chi-square	Dittman et al. (2010)
			ANOVA	Kumar et al. (2015)
			CFS	Al-Batah et al. (2019)
			FCFS	Yu and Liu (2003)
			WGCNA	Langfelder and Horvath (2008)
			PCA	Pochet et al. (2004)
			ICA	Zheng et al. (2006)
			LDA	Sharma et al. (2014)
Wrapper	Interaction between selected features and learning model taken into account. Considers feature dependencies.	Higher risk of overfitting compared to filter approach. Computationally intensive.	SFS	Park et al. (2007)
			SBE	Dhote et al. (2015)
			RFE	Guyon et al. (2002)
			GA	Ram and Kuila (2019)
			ABC	Li et al. (2016)
			ACO	Alshamlan et al. (2016)
			PSO	Sahu and Mishra (2012)
Embedded	Requires less computation than wrapper methods.	Very specific to learning technique.	k-means clustering	Aydadenta and Adiwijaya (2018)
			LASSO	Tibshiranit (1996)
			GLASSO	Meier et al. (2008)
			SGLASSO	Ma et al. (2007)
			AE	Danaee et al. (2017)
			RF	Díaz-Uriarte and Alvarez de Andrés (2006)
Hybrid	Combines filter and wrapper methods. Reduces the risk of overfitting. Lower error rate.	Computationally expensive. Can be less accurate: the filter and the wrapper both being used in different steps.	SVM-RFE	Guyon et al. (2002)
			MIMAGA-Selection	Lu et al. (2017)
			Co-ABC	Alshamlan (2018)