Metrics for Graph Partition by Using Machine Learning Techniques

Abstract

In our previous work, we explored the possibility of applying machine learning technique to graph partition. We use some metrics to describe the graph, rank the execution time of some graph algorithm and feed them into the machine learning models. We proved that decision tree and KNN and good models of this problem. In the paper, we go on to investigate more metrics to describe the graph after partitioning. We found that AverageDegreeNotCut is also an important metric. We improve the precision score of original machine learning models by 4.9 percent.