Social network analysis with prior knowledge and non-negative tensor factorization
View Patent ↗Systems and methods are disclosed to analyze a social network by generating a data tensor from social networking data; applying a non-negative tensor factorization (NTF) with user prior knowledge and preferences to generate a core tensor and facet matrices; and rendering information to social networking users based on the core tensor and facet matrices.
1. A computer-implemented method to analyze a social network, comprising
a. generating a data tensor from social networking data;
b. applying a non-negative tensor factorization (NTF) with user prior knowledge and preferences to jointly extract a core tensor and facet matrices; and
c. rendering information about the social network based on the core tensor and facet matrices, and
d. providing a plurality of levels of user inputs for each dimension of the data, including unconstrained, basis-constrained, and constant.
2. The method of claim 1 , comprising transforming the data into a high-dimensional tensor, where each dimension of the tensor corresponds to one aspect of the data.
3. The method of claim 1 , wherein each entry of the data tensor corresponds to an observed piece of data in the social network.
4. The method of claim 3 , wherein an index for a dimension is unordered or ordered.
5. The method of claim 3 , wherein each data tensor entry represents an intensity of the corresponding entry in the social networking data with a fixed set of indices in the dimension.
6. The method of claim 1 , comprising applying the NTF for parameter inference and maximum likelihood estimation.
7. The method of claim 1 , comprising incorporating prior knowledge at different levels into the factorization to control the facets in different dimensions of the data at different levels.
8. The method of claim 1 , wherein the user input level is set independently for each dimension of data.
9. The method of claim 1 , comprising applying a Dirichlet prior in a model parameter inference.
10. The method of claim 9 , comprising controlling a sparseness or smoothness of the NTF by using special Dirichlet priors.
11. The method of claim 1 , applying the NTF to select one or more communities in the social network.
12. The method of claim 1 , comprising applying the NTF to select one or more topics in the social network.
13. The method of claim 1 , comprising applying the NTF to detect one or more temporal trends in the social network.
14. The method of claim 1 , comprising applying the NTF to organizing, searching, classifying, clustering, or visualizing data in the social network.
15. The method of claim 1 , comprising generating a mode[X B 1 X 1 , . . . , X B N X N ], where X B 1 X 1 , . . . , X B N X N are the first, second, . . . , and N-th dimensions, C is the core tensor to capture the correlation among the factors, X B 1 X 1 , . . . , X B N X N .
16. The method of claim 15 , wherein X B 1 , . . . , X B N are nonnegative and encode user prior knowledge.
17. The method of claim 15 , wherein X B 1 , . . . , X B N are nonnegative.
18. The method of claim 15 , wherein a loss between the data tensor and [C, X B 1 X 1 , . . . , X B N X N ] comprises Kullback-Leibler divergence.
19. The method of claim 1 , wherein the data tensor comprises three or more dimensions.
20. A computer-implemented method to analyze a social network, comprising
a. generating a data tensor from social networking data;
b. applying a non-negative tensor factorization (NTF) with user prior knowledge and preferences to jointly extract a core tensor and facet matrices; and
c. rendering information about the social network based on the core tensor and facet matrices, and
d. generating a mode[X B 1 X 1 , . . . , X B N X N ], where X B 1 X 1 , . . . , X B N X N are the factors in first, second, . . . , and N-th dimensions, C is the core tensor to capture the correlation among the factors, X B 1 X 1 , . . . , X B N X N .