Feature Reduction Based on Genetic Algorithm and Hybrid Model for Opinion Mining

Table 4

Classification performance of DVD domain review (unigram).

Class/method	NB			LR			SVM			BSVM			BNB
	Actual neg.	Actual pos.	Acc. (%)	Actual neg.	Actual pos.	Acc. (%)	Actual neg.	Actual pos.	Acc. (%)	Actual neg.	Actual pos.	Acc. (%)	Actual neg.	Actual pos.	Acc. (%)
Unigram
Predicted neg.	82	21	82.00	86	22	86.00	75	21	75.00	86	20	86.00	88	17	88.00
Predicted pos.	18	78	78.78	14	77	77.77	25	78	78.78	14	79	79.79	12	82	82.82
Avg. accuracy	80.42%			81.95%			76.87%			82.92%			85.39% (↑)
Error rate	19.59%			18.09%			23.11%			17.08%			14.57% (↓)
Unigram + bigram
Predicted neg.	81	22	81.00	91	33	91.00	79	23	79.00	80	20	80.00	86	19	86.00
Predicted pos.	19	77	77.77	09	66	66.66	21	76	76.76	20	79	79.79	14	80	80.80
Avg. accuracy	79.42%			78.29%			77.89%			79.92%			83.47% (↑)
Error rate	20.60%			21.10%			22.11%			20.10%			16.58% (↓)