Boosting existing shortest path algorithms through highly efficient building of node cut set-based overlay

The shortest path problem is one of the fundamental problems in network optimization. The acceleration principle based on boundary reduction is implicit in the problem. However, due to the high complexity of enumeration boundaries, existing algorithms suffer from high computational complexity and only work well for small graphs, or rely on specific application knowledge and are not generally applicable. To solve this problem, we propose an efficient node cut set -based shortest path boosting algorithm (CSPA). In CSPA, the optimal and near -optimal sets of boundary nodes are efficiently located in parallel, and then a high-level tree is constructed to organize subgraphs separated by these node cut sets. The tree can be used to focus on the necessary subgraphs, split any existing preprocessing or calculation into independent and parallel operations, and then speed up any existing algorithms through problem reduction and ready -to -use parallelization. When compared with existing algorithms in random and real -world graphs, CSPA shows perfect acceleration potential with significantly lower preprocessing overhead, i.e., more than 3 orders of magnitude times faster at most. It can be used as a candidate method for shortest path acceleration in large-scale graphs.