Ordered matrix representation supporting the visual analysis of associated data
Chen, Yi; Lv, Cheng; Li, Yue; Chen, Wei; Ma, Kwan-Liu
Sci China Inf Sci, 2020, 63(8): 184101
Associated data, which refer to sets of entities with specific relations and relational weights that can usually be expressed in a relational matrix, are found in many fields. Two typical examples are the pesticide residue dataset in food-safety, in which pesticides are associated with agricultural products, and E-transaction dataset, in which there is special relation between buyer and seller. However, on large data scales, finding the key entities or mine hidden patterns in such data is difficult and time consuming. An ordered matrix can help analysts quickly locate the entity of interest. When ranking the entities, the relations and their weights should both be considered. However, the existing ranking algorithms such as PageRank, consider only the relations while ignoring their weights. In this paper, we propose a ranking algorithm called RW-Rank, in which RW stands for both relations and weights. RW-Rank is inspired by the PageRank algorithm and is available for create ordered relational matrices. Emergent visual analysis technology can improve the efficiency of complex associated-data analysis. Using this new technology, we design and implement a visual analysis system called Rank-Vis for analysing associated data by the RW-Rank algorithm.