Correct code for imputation with fancyimpute

Question

I was performing an imputation of missing values by KNN with this code:

1) data[missing] = KNN(k = 3, verbose = False).fit_transform(data[missing])

However, I saw some tutorials (e.g. Chris Albon - Machine Learning With Python Cookbook p. 78) that used the method .complete();

2) features_knn_imputed = KNN(k=5, verbose=0).complete(X_train[true_nums])

I was wondering if 2) is deprecated code or if my implementation of KNN for imputing in 1) is incorrect?

could you share the tutorial you were following so we can check it? a quick look at the fancyimpute repo indicates that KNN does not have a method called `complete()` — Gonzalo Hernandez, Oct 03 '19 at 06:38

score 1 · Accepted Answer · edited Aug 05 '20 at 13:27

1

Yes, .complete is deprecated. Use .fit_transform just like sklearn.

edited Aug 05 '20 at 13:27

Ann Kilzer

answered Aug 05 '20 at 11:41

Sea

1 Answers1