1. 默认的 KNeighborRegressor 应该能够处理回归缺失值，但是，只能处理数值.因此，对于分类值，我相信您很可能需要先对其进行编码，然后再估算缺失值.

1. The default KNeighborRegressor is supposed to be able to work with regressing missing values, however, with numeric values only. Therefore for categorical value, I believe you most likely need to encode it first, then impute the missing values.

KNNImpute，很可能使用均值/模式等

KNNImpute, most likely uses mean/mode etc

来自sklearn的iterativeimputer可以对整个数据集进行插补

iterativeimputer from sklearn can run the imputation against the whole datasets