The relationship between data can be visualized by using multiple types of images, so it is convenient for users to obtain information and relationships between data. However, when many data items, attributes and links, and the relationships between data are not clear, a visual view which is capable of relationship mining is required. For the real task requirements of domain data analysis, elements such as circle layout and node link layout were used to display simple relationships and hierarchical structures between data. What’s more, combining concentric circle layout, scatter plots, thermogram elements and dynamic filtering and data clustering techniques, a relational mining view was proposed that not only demonstrated the nature of data nodes, but also revealed the potential relationships between data. Finally, combining the above views, a visual analysis graph of the mining data relationship was presented, which was ExploreView. It was applied to the sampling data set of the Food and Drug Administration, while using cube metaphor to organize data. The bipartite graph defined task requirements before completing visual coding. It can display the basic situation of data information and dynamically interact according to the actual needs of users, and reflect the attributes, various hierarchical structures and relationships between data. As a result, the visual analysis graph was easy and efficient to operate. It can be used to provide early warning for possible food safety incidents, locate key regulatory targets, which provided reference for the development of rules, and effectively met the needs of different types of users.