Raven
Gao

This is a hierarchical interactive visualization system with specific properties, also building reflective visualization with model adjustment and filtering functions.

Horizontal version:
Heatmap version:
Publications
Gao, R., Li M., Hu, X. & Chen Y. A Hierarchical Interaction Design for Multi-dimensional Flow Datasets. Poster presented at: IEEE Visual Analytics Science and Technology (VAST), IEEE VIS 2016; 2016 Oct. 22-28; Baltimore, MD.

1. Selector View:
A heatmap of the number of patterns based on different parameters such as support (x-aixs, proportion of pattern existence in all), and aiScore (ratio of active / inactive the pattern locate in) Once you select the specific range, it will generate a list of patterns that satisfy the conditions. View 2 of parallel coordinate updated.

2. Parallel Coordinate + Network:
The dimensions of all patterns selected include: support, aiScore, active compound #, inactive compound # and number of elements in each pattern. Each 2 dimensions (parameter) will form scatter plot for further analysis. Interaction include brushing on both parallel coordinate and scatter plot. The it will generate another list of patterns selected on the up left corner. You can then click on one specific pattern to see more information. go to view 3.

Tree network is a hierarchical network of selected pattern. You can choose confidence level to generate new patterns. (e.g.: pattern 134 will generate patterns like 1345, 1347, 1349) You can click on each node to either collapse or expand children nodes. Color of each node indicate the active vs inactive compound ratio. Size is the frequency of the pattern.

This is a visualization system showing features and network to discriminate patterns of compounds and inner substructures based on medical data.

This system aims to present a matrix-based view to present different type of relationships and data feature presentation. For compound- structure view, For inner relationship exploration of compounds or structures, we can use collaborative filtering to construct self-to-self relationship with the compound-structure matrix we already have. A network can be represented by an adjacency matrix, where each cell ij represents an edge from vertex i to vertex j. Here, vertices represent relationship compounds or structures (in a compound), while edges represent co-occurrence value.

The data basically consists of one large table of compound and the presence and absence of substructures in each compound. Some compounds are active, others are not. The objective is to extract/visualize set of bits “on” which discriminate the actives from the inactives. As this file is very large, we usually compress it. So each column now corresponds to several substructures as shown in the second table. Then a value 1 means that at least one of the substructure is present, and a value of 0 means that none of the substructure is present in the fingerprint.

We present a system of multiple interactive and coordinated views to help visualize and analyze several kinds of large-scale data and complex datasets. This design assists to explore social network involved and analyze natural language used in large-scale datasets, then to make an integrated analysis and presentation of all kinds of objects involved in cases in multiple perspectives and hierarchies. We implemented datasets from VAST Challenge 2014 mini challenge 1 to verify the usability of our design.

Click on pictures below to interact with the graph

1. Coordinate Graph View

2. Multi-lane timeline & social network linkage view

3. Cluster analysis view

4. Event timeline view

Video Demo:

Award & Publication
Vast Challenge 2014: Honorable Mention for Effective Use of Coordinated Visualizations
Gao, R., Tao, H., Chen, H., Wang, W., & Zhang, J. (2014, October). Multi-view display coordinated visualization design for crime solving analysis: Vast challenge 2014: Honorable mention for effective use of coordinated visualizations. In Visual Analytics Science and Technology (VAST), 2014 IEEE Conference on (pp. 321-322). IEEE.