An input variables selection method is proposed based on the k-nearest neighbors mutual information. According
to the mutual information between the multi-variables, the relevant variables which have great influence to the mutual
information are selected. Meanwhile, the redundant variables are removed according to the mutual information between
the input variables sets. Consequently, the input variables are selected with proper parameter settings. The results of the
simulation based on the Friedman data, the Lorenz time series and the Housing data show the effectiveness of the proposed
input variables selection method.