Mining program workflow from interleaved traces

摘要: Successful software maintenance is becoming increasingly critical due to the increasing dependence of our society and economy on systems. One key problem difficulty in understanding evolving Program workflows can help system operators administrators understand behaviors verify executions so as greatly facilitate maintenance. In this paper, we propose an algorithm automatically discover program from event traces that record events during execution. Different existing workflow mining algorithms, approach construct concurrent interleaved events. Our a three-step coarse-to-fine algorithm. At first, mine temporal dependencies for each pair Then, based mined pair-wise tem-poral dependencies, basic model by breadth-first path pruning After that, refine verifying it with all training traces. The re-finement tries find out interpret minimal state transitions threads. results both simulation data real show highly effective.