We present a new technique for discovering and representing structural changes between two versions of a directed acyclic graph (DAG). Motivated by the necessity of change detection in temporal data warehouses and inspired by a well known tree comparison algorithm, we developed a heuristic method to calculate an edit script transforming an old version of a graph to the new one. This edit script is composed of operations for inserting and deleting nodes and changing labels and values of nodes as well as for inserting and deleting edges to cover rearrangements of nodes (moves). We present the prerequisites of our approach, the different phases of the algorithm and discuss some evaluation results gained from a prototypic implementation. Our approach is applicable to arbitrary labeled DAGs in any context, but optimized for rooted, ordered and labeled acyclic digraphs with a small rate of changes between the DAGs to be compared.

Top