In a dataset derived from census information, which comprises several correlated feature columns, the column indicating a person's age is incomplete, with 10% of its values missing. Considering the goal of maximizing the predictive accuracy of a model developed using this dataset, which strategy should be employed to address the missing values in the age column?