The Feature F-measure is a statistical and parameter-free metric used in feature selection that performs well for classification, clustering, cluster labeling, also used to evaluate cluster quality. We evaluate its use in the complex networks framework. We are especially interested in evaluating its use to characterize the node connectivity regarding the community structure. This would allow to benefit from its parameter-free system of feature selection, and of its well-evaluated performance. We thus study on a benchmark of realistic synthetic graphs the correlations between Feature F-measure and classic centrality measures, but also between Feature F-measure and measures designed to characterize community roles of nodes. We show that Feature F-Measure is linked to node centrality, and that it is well-fitted to evaluate their connectivity w.r.t. the community structure. We also observe that community roles detection measures are dependent of the community size, whereas Feature F-Measure is tied to density, which makes results comparable from a network to another. This allows to consider using Feature F-Measure to study dynamic temporal network, using it for community matching.