A genetic algorithm for discriminative graph pattern mining

In this paper, we introduce a new algorithm EWALDIS (Evolution- and random WALk-based algorihm for DIScriminative patterns) for mining discriminative patterns on the local level of dynamic attributed multigraphs. It uses a random walk-based approach [1] and a genetic algorithm to mine inexact patterns from the perspective of attributes and also times-tamps. This also means that it does not require the discretization of the timestamps to be able to find some patterns. Moreover, by utilizing sampling techniques, the algorithm does not have to traverse the whole search space. EWALDIS is an improved version of WALDIS [7].