A Computational Approach to Improving Fairness in K-means Clustering