Better Customer Clustering Using an Initialization Approach for the K-Means Algorithm