The invention provides a network unknown connection edge prediction method based on a second-order
local community and preferential attachment. The method comprises the following steps: constructing a
network model, and obtaining first-order
common neighbor nodes and second-order
common neighbor nodes of a pair of unconnected nodes, wherein the nodes and the connection edges between the nodes form the second-order
local community; recording the total number of the nodes and connection edges of the
local community, and meanwhile, recording the number of neighbors, outside the
community, of two seed nodes; calculating volume coefficient, edge-
clustering coefficient and simple
harmonic average distance of the
community and second-order local
community coefficient; calculating similarity
score index between the node pair; traversing the whole network, and for any two unconnected nodes, calculating similarity
score index between the corresponding node pair; and
ranking similarity scores between all of the unconnected node pairs in a descending order, and taking nodes corresponding to the front m
score indexes as predicted connection edges. The method takes the second-order local community and preferential attachment information of the seed nodes into consideration, and makes full utilization of information of a network
local structure, and is good in prediction effect and high in accuracy.