ABSTRACT
ABSTRACT: In complex network graph, connectivity of graph depends on the second smallest eigenvalue of the Laplacian matrix. Correlation coefficients are introduced to this cut model to resolve overlapping community detection by minimizing the algebraic connectivity of complex networks. In this paper, we define edge centrality for each edge by spectral analysis and propose an advanced algorithm of community detection on the basis of centrality measure and correlation coefficients. By the analysis of the algorithm and missing value processing, three methods of missing value processing are put forward. The study subject is classified into several groups by the algorithm of community detection on the basis of centrality measure. Then, it calculates the centrality of the study subject with missing value in the group and deletes the record with a low centrality; on the contrary, the traditional methods of missing value processing are adapted to process missing data in same group. This method is applied to evaluate a fast-food company with missing data. The results show that the missing value processing method based on community detected outperforms the traditional mean imputation method, multiple imputation method, and K-means algorithm. It offers a practical approach for missing value processing.
