SizePairs: Achieving Stable and Balanced Temporal Treemaps using
Hierarchical Size-based Pairing

(Proceedings of IEEE VIS 2022)

Chang Han 1    Jaemin Jo 2    Anyi Li1    Bongshin Lee 3    Oliver Deussen 4    Yunhai Wang 1  
1Shandong University   2Sungkyunkwan University   3Microsoft Research   4Konstanz University, Germany  

Figure 1: Comparing different temporal treemap methods using three time steps (t = 0, t = 10, and t = 20) of the WorldBankHIV data: (a) Hilbert treemap method [26]; (b) Local moves method [25]; (c) Greedy insertion treemaps [29]; (d) SizePairs; and (e) SizePairs with local moves. To reflect the stability of each method, we assign colors to each rectangle (based on its position) at the first time step and then apply the color scheme to the other time steps (we mark three data items using shpaes to facilitate the comparison). SizePairs is more stable and maintains better aspect ratios, and SizePairs in combination with local moves (e) has even better aspect ratios than SizePairs alone (d).


We present SizePairs, a new technique to create stable and balanced treemap layouts that visualize values changing over time in hierarchical data. To achieve an overall high-quality result across all time steps in terms of stability and aspect ratio, SizePairs employs a new hierarchical size-based pairing algorithm that recursively pairs two nodes that complement their size changes over time and have similar sizes. SizePairs maximizes the visual quality and stability by optimizing the splitting orientation of each internal node and flipping leaf nodes, if necessary. We also present a comprehensive comparison of SizePairs against the state-of-the-art treemaps developed for visualizing time-dependent data. SizePairs outperforms existing techniques in both visual quality and stability, while being faster than the local moves technique.

Source Code:



Paper: [PDF 1.6M].


Figure 2: Treemaps for the NetMigration dataset generated by SP-F, LM4, and GIT. Nodes that disappear in 2012 and re-appear in 2013 are placed consistently to their previous locations by SP-F.

Figure 3: Relationship between visual quality (aspect ratio) and short-term stability (corner-travel distance) of different methods for datasets grouped by their characteristics. Note that for aspect ratio, a higher value is better, and for stability, a lower value is better. Therefore, a method is superior to another if its point is closer to the bottom-right corner.

Figure 4: Superiority of SP(F) over LM and GIT. Green cells indicate SP(F) outperforms GIT or LM on average in terms of corner-travel distance and aspect ratio for datasets with a certain feature combination. Gray means there is a trade-off (e.g., aspect ratio is better but stability is worse). Red means our method was inferior in terms of both metrics.


This work was supported by the grants of the NSFC (62132017, 62141217), and the grants of National Research Foundation of Korea (No. NRF-2021R1F1A1055995).