Suppose I have a dataset consists of 8 features except target feature. Each have different number of missing value. Which feature should I impute first? Feature with lowest missing value or in the contrast? Is it make different? In case I use machine learning like decision tree or linear regression to impute missing value can I use feature with NA to impute another feature.
You can follow these steps,
- First, if you can logically impute missing values in any feature, do that.
- If there are missing values still in every feature, drop the rows specifically which are missing in lowest missing values column and repeat step 1 again.
- Then, use methods mentioned here for imputing missing values
For your last question, yes you can use ML methods.