Table 3 Predictive performance of machine learning models with variable selection algorithms for peroxide and iodine value prediction in crude palm oil

Parameter	Machine Learning Models	Number of variables	Calibration set n = 133		Prediction set n = 67		RPD
Parameter	Machine Learning Models	Number of variables	Rc	RMSEC	Rp	RMSEP	RPD
Peroxide value (meq O₂/kg)	SVM	1024	0.9796	0.3130	0.9439	0.8715	2.1877
	CARS- SVM	104	0.9799	0.2998	0.9584	0.8146	2.6223
	UVE- SVM	66	0.9801	0.3000	0.9736	0.7596	2.7392
	GA- SVM	148	0.9761	0.3272	0.9065	0.6531	3.0566
	RF	1024	0.9833	0.1459	0.9802	0.2735	6.2240
	CARS-RF	104	0.9916	0.1928	0.9599	0.4252	5.0331
	UVE-RF	66	0.9868	0.2418	0.9589	0.4304	4.9724
	GA-RF	148	0.9959	0.1346	0.9831	0.2765	7.7397
Iodine value (g I₂/100 g)	SVM	1024	0.9807	0.3615	0.9559	0.9236	2.3871
	CARS- SVM	91	0.9789	0.3306	0.9424	0.9784	2.4697
	UVE- SVM	45	0.9598	0.4545	0.9125	0.6965	3.4467
	GA- SVM	140	0.9787	0.3367	0.9030	0.7103	2.9234
	RF	1024	0.9887	0.2393	0.9725	0.3748	6.0727
	CARS-RF	91	0.9888	0.2395	0.9475	0.5179	4.3962
	UVE-RF	45	0.9842	0.2842	0.9678	0.4058	5.6106
	GA-RF	140	0.9947	0.1638	0.9752	0.3561	6.3927

n number of samples, Rc Correlation Coefficient for the calibration set, RMSEC root mean square error of cross-validation Rp correlation coefficient for the predictionn set, RMSEP root mean square error of prediction, RPD residual predicted deviation, SVM support vector machine, CARS-SVM competitive adaptive reweighted sampling - support vector machine, UVE-SVM uninformative variable elimination - support vector machine, GA-SVM genetic algorithm - support vector machine, RF random forest, CARS-RF competitive adaptive reweighted sampling - random forest, UVE-RF uninformative variable elimination - random forest, GA-RF genetic algorithm - random forest.

Quick links

Search