Refrain from imputing the missing values, allowing the machine learning algorithm to decide how to handle them.

Eliminate all feature variables that initially contained missing values from the feature set.

Impute the missing values using the mean value of each respective feature variable instead of the median.

Create a constant feature variable for each feature with missing values, showing the percentage of rows originally missing from the feature.