Front-Page News Classification Model Based on the Stacking of Textual Context and Attribute Information

Table 3

The average values of four evaluation indexes in respect to 7 classification models. In the bracket, four evaluation indexes are increasing or declining degrees in comparison of the classification results in Table 2 (the top 2 performers are highlighted in bold).

Index	XGBoost	RF	FastText	TextCNN	TextRNN	TextRCNN	StacText
Specificity	0.832 (−15.4%)	0.998 (0.8%)	0.710 (−28.0%)	0.784 (−19.5%)	0.762 (−22.6%)	0.799 (−18.4%)	0.942 (−2.9%)
Precision	0.331 (−46.1%)	0.574 (88.2%)	0.261 (−53.1%)	0.287 (−48.8%)	0.272 (−43.1%)	0.333 (−47.3%)	0.541 (−11.7%
Recall	0.595 (205.1%)	0.018 (−40.0%)	0.667 (240.3%)	0.602 (162.9%)	0.647 (151.8%)	0.593 (110.3%)	0.447 (42.8%)
-score	0.424 (43.2%)	0.034 (−37.0%)	0.362 (26.1%)	0.377 (4.7%)	0.375 (2.8%)	0.397 (1.8%)	0.482 (16.7%)