Thank you for your all comments and help.
I have searched the issue for few days,
I have read some articles and find good and detailed explanation of which graphic type I can use. I want share the information with you.
"....When the co-occurrence graph is directed, it means that
one type of entity is responsible for generating another type,
therefore it is logical to define the generative model as a
conditional probability. In the case of an undirected graph,
the joint probability is more appropriate.
Based on the categories for the co-occurrence data, four
different classes can be defined. In hetero-directed, each token
consists of two different entities and one type of entity is
responsible for generating the other. The most popular example
of this type is text data where a document is responsible
for generating words. Therefore, the link direction is from
document nodes to word nodes. In hetero-undirected, each
token consists of two different entities and both entities
are generated simultaneously from a joint distribution. An
example of this type is the co-occurrence of image features
Mohammad Khoshneshin, W. Nick Street, and Padmini Srinivasan. Bayesian embedding of co-occurrence data
for query-based visualization. In Proceedings of IEEE International Conference on Machine Learning and