Complex networks permeate our daily lives. Their structures play a central role in social, biological, and economic interactions. By improving the accuracy of modeling network structures, we would achieve unprecedented improvements in controlling the destructive impacts of negative networks; such as criminal or cancer networks, or extending the scale of a good behavior in a complex network. Expectation-Maximization technique is applied along with maximum likelihood technique to improve the degree distribution estimates and, in the end, improve the accuracy of network structures. In addition to the networks' static structure, i.e. degree distribution, this study evaluates the network's structural change under random failures and targeted attacks. Removing nodes from a network is another challenge in complex network area. Six network performance metrics are introduced to quantify diffusion speed, diffusion scale, homogeneity, and diameter of a network. Moreover, ten heuristic node removal strategies are designed using different node centrality metrics including degree, betweenness, reciprocal closeness, complement-derived closeness, and eigenvector centrality. The experiments find that degree and betweenness could select more important nodes than the other node centrality metrics. Since calculating degree is computationally more reasonable than betweenness, we use this metric to remove nodes from September 11 hijackers' network, HIV network in Colorado Springs, and the power grid network in North America.