An Earthquake Emergency Web Data Cleaning and Classification Method Based on Word Frequency and Position Weighting

Table 6

Model training result efficiency compared.

N-gram model (N)	Number of gram	Feature factor selection in P-TF-IDF algorithm	Accuracy	Training time (s)	P-TF-IDF after filtering enter percentage of words (%)	Word count
N = 2 (Bi-gram)	2.2 ∗ 10^9	Add word frequency factor and location factor	0.9026	27.52	25	47071
N = 3 (Tri -gram)	1.0 ∗ 10^14	Add word frequency factor and location factor	0.9015	34.47
N = 4 (4-gram)	4.9 ∗ 10^18	Add word frequency factor and location factor	0.9015	47.17
N = 5 (5-gram)	2.3 ∗ 10^23	Add word frequency factor and location factor	0.8997	52.55
N = 2 (Bi-gram)	8.4 ∗ 10^9	Add word frequency factor and location factor	0.9097	20.69	50	91711
N = 3 (Tri -gram)	7.7 ∗ 10^14	Add word frequency factor and location factor	0.9073	31.22
N = 4 (4-gram)	7.0 ∗ 10^19	Add word frequency factor and location factor	0.9056	33.64
N = 5 (5-gram)	6.4 ∗ 10^24	Add word frequency factor and location factor	0.9038	45.79
N = 2 (Bi-gram)	1.7 ∗ 10^15	Add word frequency factor and location factor	0.9021	23.52	75	132660
N = 3 (Tri -gram)	2.3 ∗	Add word frequency factor and location factor	0.9021	36.67
N = 4 (4-gram)	3.0 ∗	Add word frequency factor and location factor	0.9003	48.61
N = 5 (5-gram)	4.1 ∗	Add word frequency factor and location factor	0.8985	58.81
N = 2 (Bi-gram)	3.6 ∗	None	0.9021	28.26	100	191297
N = 3 (Tri -gram)	7.0 ∗	None	0.9011	36.31
N = 4 (4-gram)	1.3 ∗	None	0.9001	49.59
N = 5 (5-gram)	2.5 ∗ 10^26	None	0.8937	68.58