Evolutionary Data Preprocessing to Alleviate Class Imbalance

Table 6

Class imbalance ratios of the training, validation, and test datasets.

Classes		Train and validation	Imbalance ratio (%)	Test	Imbalance ratio (%)
1	Normal	23,112	2.81	739,588	43.40
2	Exploits	14,842	1.78	14,841	24.14
3	Reconnaissance	4,662	0.55	4,663	6.50
4	DoS	5,451	0.65	5,451	7.69
5	Generic	17,957	2.17	71,828	30.76
6	Shellcode	504	0.06	504	0.66
7	Fuzzers	8,082	0.96	8,082	11.84
8	Worms	58	0.01	58	0.08
9	Backdoor	776	0.09	776	1.03
10	Analysis	893	0.11	892	1.18
Sum		76,337	1.00	846,683	142.35
Standard deviation		8,250	10.19	231,133	15.05