3 ways to handle missing values in Machine Learning
Steps to handle missing values:
Drop the columns that has missing values
Here, we drop the columns that have missing values.
With this approach, there is a chance that we may lose access to lot of useful information.
You should use this approach only if most of the values in a column are missing and not otherwise.
Imputation
Here, we replace all the missing values in a column with the average value of that column. For example, if column has 3, 2, 1 as values then we take average of this column that is 2 and replace it with missing values.
It’s a standard approach.
The imputed value is not always right, however, it results in good accuracy as compared to the previous way.
Extension to Imputation
Here, we follow the process of imputation and then for each column with missing value we also create a new column.
The location at which the value is imputed we mark an entry in the newly created column to specify that this location has imputed value.
This may increase the accuracy of prediction, but in some cases it can also make no difference.
