Parametric and Nonparametric Approaches of Reid Vapor Pressure Prediction for Gasoline Containing Oxygenates: A Comparative Analysis Using Partial Least Squares, Nonlinear, and LOWESS Regression Modelling Strategies with Physical Properties

Table 3

Accuracy of PLSR, NLR, and NPR models for predicting gasoline RVP and comparison with previous studies.

Investigation approach	MAPE (%)	MAD (kPa)	RMSE (kPa)	No. of samples	No. of input variables
PLSR (this work)	7.752	4.131	5.321	904	6
NLR (this work)	7.661	4.096	5.253	904	6
NPR (LOWESS) (this work)	7.020	3.933	5.163	904	6
PC-SAFT^a	4.030			7	80
PSRK^a	7.610			7	80
Wilson Eq.-LS^b	5.640			24	24
Wilson Eq.-LI^b	4.410			24	24
UNIFAC^c	1.0–10.0			5	16
LSSVM^d (without oxygenates)			9.231	362	23
ANN^e (without oxygenates)		2.280		362	27
SAFT-γ EoS^f	4.36-10.59			7	18
CPA^g	5.500			4	226

^aPC-SAFT is perturbed-chain statistical associating fluid theory, and PSRK is predictive Soave-Redlich-Kwong equation adopted from Vella and Marshall [56], and they used data of gasoline and methanol blend taken from the experimental work of Andersen et al. [57]; ^bLI is Lagrange interpolating polynomial statistical method, and LS is least squares statistical fitting method adopted from Pumphrey et al. [58] for gasoline and isopropanol blend samples; ^cUNIFAC is universal quasichemical functional group activity coefficients adopted from Hatzioannidis et al. [59] for gasoline containing MTBE, methanol, ethanol, and isopropanol blends; ^dLSSVM is least squares support vector machine adopted from Kamari et al. [14] (considered for gasoline only); ^eANN is artificial neural network adopted from Albahri et al. [60] and considered for gasoline only; ^fSAFT-γ EoS is statistical associating fluid theory–Mie model equation of state adopted from Landera et al. [61]; ^gCPA is cubic plus association model adopted from Gaspar et al. [7].