Using ESDA to Detect Overlapping Multi-communities

Traditional algorithm in Community identification take full advantage of vertex. But in these algorithms, node´s aggregation characteristics are not obvious and the quantity of communities is not reasonable. The edge of the spectral decomposition algorithm (ESDA) is different from traditional method for community partition. There are four steps in ESDA: first, we translate the origin graph into line graph. Second, edge degree for 1 and the special local gathered structure are dealt by pre-processing to simplify complex networks. Third, ESDA would use the second smallest, third smallest, the special eigenvalue corresponding to eigenvector to build up coordinate system. Finally, we can identify community by using coordinate system. Experiments show that this algorithm not only make more prominent characteristics of community together and has a better effect, but also speeds up partition of community by sub-step pretreatment.