Table 1.

Overview of different ML-based methods utilized in the previous literature for diabetes prediction, including the year of publication, used dataset, missing value imputation techniques, feature selection strategies, number of selected features, classifier used, and corresponding performance evaluation metrics.

Years	Dataset	MVI 1	FS	NSF	BPC	Performance
2016 [32]	ENRC	None	None	9	DT	$Acc$: 0.840
2018 [37]	LMHC	None	None	All	RF	$Acc$: 0.808 $S_n$: 0.849 $S_p$: 0.767
2018 [37]	PIDD	None	mRMR	7	RF	$Acc$: 0.772 $S_n$: 0.746 $S_p$: 0.799
2018 [30]	PIDD	None	None	8	NB	$AUC$: 0.819 $Acc$: 0.763 $S_n$: 0.763
2018 [38]	PIDD	KNN impute	BWA	4	Linear Kernel SVM	$AUC$: 0.920
2019 [39]	PIDD	NB	None	8	RF	$AUC$: 0.928 $Acc$: 0.871 $S_n$: 0.857
2019 [40]	PIDD	None	CRB	11	NB	$Acc$: 0.823
2019 [41]	PIDD	None	None	8	MLP	$Acc$: 0.775 $S_n$: 0.85 $S_p$: 0.68
2020 [31]	PIDD	Mean	CRB	6	Ensemble of AB, XGB	$AUC$: 0.950 $S_n$: 0.789 $S_p$: 0.789
2020 [42]	NHANES	None	LR	7	RF	$AUC$: 0.95 $Acc$: 0.943
2020 [43]	PIDD	Case deletion	None	2	SVM	$AUC$: 0.700 $Acc$: 0.750
2021 [44]	PIDD	None	None	8	Ensemble of J48, NBT, RF, Simple CART, RT	$AUC$: 0.832 $Acc$: 0.792 $S_n$: 0.786
2021 [45]	LMHC	Case deletion	ANOVA, GI	16	XGB	$AUC$: 0.876 $Acc$: 0.727 $S_n$: 0.738

¹ Note: MVI: Missing Value Imputation, FS: Feature Selection, NSF: Number of Selected Feature, BPC: Best Performing Classifier, ENRC: Egyptian National Research Center, LMHC: Luzhou Municipal Health Commission, PIDD: PIMA Indian Dataset, mRMR: Minimum Redundancy Maximum Relevance, BWA: Boruta Wrapper Algorithm, CRB: Correlation-Based, NHANES: National Health and Nutrition Examination Survey, ANOVA: Analysis of Variance, GI: Gini Impurity, NBT: Naive Bayes Tree, RT: Random Tree.