Detecting Web Spam Based on Novel Features from Web Page Source Code

Table 6

Logit regression results.

Type	Feature	Coefficients	Std. error	Pr(>—z—)
Selected existing features	HST_19	0.7254	0.338	0.032
	HST_20	2.3106	0.420	0.000
	outdegree_hp	0.0066	0.003	0.015
	pagerank_hp	−9.712e + 04	4.2e + 05	0.817
	truncatedpagerank_1_hp	−1.043e + 05	1.35e + 06	0.939
	truncatedpagerank_2_hp	−1.079e + 05	2.35e + 06	0.963
	truncatedpagerank_3_hp	−1.101e + 05	4.28e + 06	0.980
	truncatedpagerank_4_hp	−1.084e + 05	2.74e + 06	0.968
	L_outdegree_hp	−0.0425	0.002	0.000
	L_pagerank_hp	0.0472	0.012	0.000
	L_truncatedpagerank_1_hp	6.4183	0.665	0.000
	L_truncatedpagerank_2_hp	−8.9288	1.766	0.000
	L_truncatedpagerank_3_hp	3.5543	2.380	0.135
	L_truncatedpagerank_4_hp	−0.9848	1.369	0.472
Novel features	Diversity of HTML tags	−0.0523	0.011	0.000
	Depth of element nodes	−0.0123	0.002	0.000
	Number of HTML tags	−0.043	0.000	0.000
	Number of external links	0.0460	0.003	0.000
	Number of cross links	0.0173	0.002	0.000
	Similarity of texts	−0.0816	0.029	0.004
	Similarity of links	1.7910	0.249	0.000