Detecting anomalies in spatiotemporal data using genetic algorithms with fuzzy community membership

A genetic algorithm is combined with two variants of the modularity (Q) network analysis metric to examine a substantial amount fisheries catch data. The data set produces one of the largest networks evaluated to date by genetic algorithms applied to network community analysis. Rather than use GA to decide community structure that simply maximizes modularity of a network, as is typical, we use two fuzzy community membership functions applied to natural temporal divisions in the network so the GA is used to find interesting areas of the search space through maximization of modularity. The work examines the performance of the genetic algorithm against simulated annealing using both types of fuzzy community membership functions. The algorithms are used in an existing visualization software prototype, where the solutions are evaluated by a fisheries expert. Keywords-spatiotemporal visualization; fisheries; genetic algorithm; Q modularity; fuzzy community membership

• Type: Journal
• Year: 2010
• Where: ISDA
• Authors: Garnett Carl Wilson, Simon Harding, Orland Hoeber, Rodolphe Devillers, Wolfgang Banzhaf
• Keywords: Community Membership Functions, Fuzzy Community Membership, Genetic Algorithm, ISDA 2010, Operating System