Contextual polarity and influence mining in online social networks

Computational Social Networks

Table 7 (Disagreeable ideas) Modularity measures, detected communities, and significant communities at level 0.05 (based on the Wilcoxon rank-sum test)

Community detection algorithm	Modularity	Number of communities	Significant communities
With weight
Fast Greedy	0.44	11	10
Multi-level	0.44	9	9
Edge Betweenness	0.41	15	9
Leading Eigenvector	0.41	15	9
Info Map	0.41	34	8
Walk Trap	0.39	23	7
Spin Glass	0.25	15	8
Label Propagation	0.14	9	4
Without weight
Fast Greedy	0.47	8	8
Multi-level	0.44	9	9
Edge Betweenness	0.45	20	8
Leading Eigenvector	0.43	20	8
Info Map	0.42	35	7
Walk Trap	0.38	32	6
Spin Glass	0.25	11	8
Label Propagation	0.05	9	4

Each community structure is detected by one of the eight algorithms, with and without sentiment scores as weight