Table 2.

Complement effects of different methods on important features with missing data.

Complement methods	Average Value	Frequent Value	Specified Value	Median Value	2-k-NN	3-k-NN
Network	Extraction or Non-extraction (accuracy of complete data: 0.9404)
Crowding, Upper arch	0.8642	0.8113	0.8113	0.8742	0.8775	0.8841
Crowding, Lower arch	0.8974	0.8742	0.8709	0.8642	0.8808	0.8808
U1-NA°	0.9106	0.9073	0.8974	0.9007	0.9238	0.9172
Network	Extraction Patterns (accuracy of complete data: 0.8333)
ANB	0.8198	0.8063	0.8198	0.8153	0.8268	0.8268
Overbite	0.8198	0.8153	0.8153	0.8108	0.8288	0.8243
Lip incompetence	0.8288	0.8243	0.8243	0.6982	0.8063	0.8018
Network	Anchorage Patterns (accuracy of complete data: 0.9279)
Curve of Spee	0.8649	0.8694	0.8694	0.8649	0.8874	0.8829
Nasolabial angle	0.8784	0.8784	0.8784	0.8784	0.8468	0.8559
UL-EP	0.8739	0.8333	0.8739	0.8739	0.8694	0.8919

The highest complement accuracy for each feature is shown in bold.