Evolutionary Graph Clustering

Graph clustering is the detection of tightly connected regions in a graph. A clustering may yield structural information about the graph that is especially valuable in this data driven age. As graph clustering is considered to be an NP-hard optimization problem, the effort of accurately computing the best solution clustering for larger graphs is intractable. On these grounds, heuristics that search for an acceptable solution are used to find candidate solutions to graph clustering problems while having limited knowledge about the problem and not exceeding time limitations. Evolutionary algorithms are a type of metaheuristic optimization algorithm and search for solutions by applying mechanisms of evolution, and have been proven to be a good choice for NP-complete problems. In this thesis, we take on defining and implementing an evolutionary algorithm that aims to improve on candidate solutions found by a base algorithm. We evaluate our algorithm by comparing our results to the results achieved in a graph clustering benchmark.