In this paper I will introduce a new idea K-dijkstra. K-dijkstra is using advanced BFS (breadth rst search) and dijkstra, the partition idea is based on the idea of k-means, which takes important role in Traditional Partitioning. At the same time, K-dijkstra has a high eciency with a forceful partition result, it observes a 100x speed-up compared with Traditional Partitioning.
In order to satisfy more users requirement, a hybrid algorithm which called K-mix will also be introduced at the end. K-mix has both the high precision as Traditional Partitioning, when deal with same data set, K-mix may mostly save 15% process time compared with Traditional Partitioning; and thanks to K-dijkstra, it is much faster than Traditional Partitioning. So total three ideas are explained in the paper, users can select the favorite one to use for graph partitioning, and optimize it to a much better eectiveness.