Frequent Subgraph Miners: Runtimes Don’t Say Everything

In recent years several frequent subgraph miners were proposed. The authors of these new algorithms typically compared the runtimes of their implementations with those of previous implementations to confirm the efficiency of their methods. To get a better perspective on the mutual benefits of the algorithms, Wörlein et al. [9] performed an experimental evaluation of re-implementations of several depth-first graph miners, where also some statistics beyond runtimes were compared. In this paper we present results of an additional experimental comparison of several graph miners, which differs in the following aspects from this previous study: (1) we compare original implementations; (2) we compare these implementations on a larger set of measures than runtimes, thus providing further insight in the benefits of the algorithms; (3) we include breadth-first graph miners and free tree miners in the comparison.