Using a modified weighted K-means clustering model with custom distance to find the optimal distribution centers