How to estimate K value without domain knowledge in K-means

K-means is one of the most famous methods in classification problems. And which is used in various fields because it is easy to implement and fast to classification speed. However, since the number of clusters K to be created must be determined before clustering, it is difficult to expect good performance in data without domain knowledge. To resolve this problem, we propose RK-means (Repulsive K-means) which removes empty clusters through repulsive force between clusters. The RK-means pushes the center point of the smaller mass clusters out of the data group. This provides the opportunity for the center points to find and exploit other data groups or to be empty clusters. By eliminating empty clusters, the dataset can be partitioned into the appropriate number of clusters without domain knowledge.