According to this table, the cluster-based LLS imputation
method (CLLS) perform consistently better than the classical
KNN impute. This also holds true as compared to the two
cluster-based KNN models, i.e., CKNN and CWKNN. These
differences are obvious on Alpha, cyc.a and cyc.b, as compared
to the other three datasets. In addition, it is certain that CLLS is
generally more effective than the KM baseline over all the six
datasets. However, KM achieves reasonable results when the
level of missing value increases. As compared to the base line
model of LLS, CLLS can improve the accuracy of estimates in
all datasets examined herein. This improvement usually gets
larger with the amount of missing values. As for example,
Figs 1-3 show the focused comparisons of CLLS and LLS on
cyc.a, env and ta.crc, respectively.