Data Imputation for Detected Traffic Volume of Freeway Using Regression of Multilayer Perceptron

Table 1

Measurement conversions.

Imputation method	MAPE or RMSE	Urban road or freeway	Data sources
Principal component analysis (PCA)	76∼84 [32]	Urban road	China
	14∼24 [31]	Freeway	America
Fuzzy rough set (FRS)	4.768∼6.533 [33]	Freeway	China
	6.9766∼10.4998 [34]	Freeway	China
Tensor completion	4.0893∼5.3544 [35]	Urban road	China
	10.3%∼12.71% [36]	Freeway	America
	7.3%∼19.98% [37]	Urban road	China
	0.91%∼64.95% [38]	Urban road	China
Support vector machine (SVR)	4∼14 [40]	Both	China
	5.7232% [41]	Urban road	China
Denoising stacked autoencoders (DSAE)	13.9∼20.9 [42]	Freeway	America
Convolutional neural network (CNN)	≤24 [44]	Freeway	America
Long short-term memory (LSTM)	9.63∼17.54 [46]	Urban road	America
	1.927∼9.192 [47]	Freeway	America
Generative adversarial networks (GAN)	3.66∼10.86 [48]	Urban road	China and America
Dual-stage error-corrected boosting regressor (GBR)	1.39%∼6.08% [49]	Freeway	America
Spatiotemporal-PTD	3.45∼8.35 [50]	Urban road	China