A Comparative Analysis of Fraudulent Recruitment Advertisement Detection Methods in the IoT Environment

Table 4

Experimental results on the balanced dataset.

Category	Model	Balanced dataset
		Traditional learning	RF	0.899	0.899	0.900	0.900
SVM	0.884		0.885	0.885	0.884
LR	0.897		0.898	0.899	0.897
NB	0.893		0.894	0.894	0.893
Traditional learning + feature extraction	RF+TF-IDF	0.923	0.923	0.923	0.923
	SVM+TF-IDF	0.928	0.930	0.927	0.928
	LR+TF-IDF	0.908	0.908	0.909	0.908
	NB+TF-IDF	0.876	0.879	0.879	0.876
Deep learning	GRU	0.835	0.836	0.835	0.834
	Bi-LSTM	0.884	0.886	0.884	0.884
	TextCNN	0.930	0.931	0.930	0.930
Deep learning + pretraining	GRU+Word2Vec	0.748	0.750	0.745	0.747
	GRU+GloVe	0.748	0.748	0.748	0.748
	Bi-LSTM + Word2Vec	0.875	0.875	0.876	0.875
	Bi-LSTM + GloVe	0.831	0.832	0.830	0.830
	TextCNN + Word2Vec	0.923	0.923	0.923	0.923
	TextCNN + GloVe	0.926	0.927	0.927	0.927