You are conducting exploratory data analysis on a dataset and encounter an important categorical feature that has 5% missing values. To ensure the integrity of your analysis and to minimize any potential bias from these missing values, which of the following approaches would be the best way to handle these missing values?