IKDV-SYS: A System for Near Real-Time Interactive Kernel Density Visualization

Tsz Nam Chan1, Pak Lon Ip2,3, Leong Hou U2,3, Weng Hou Tong3, Shivansh Mittal4, Ye Li2,3 and Reynold Cheng4

1Department of Computer Science, Hong Kong Baptist University
2State Key Laboratory of Internet of Things for Smart City
3Department of Computer and Information Science, University of Macau
4Department of Computer Science, The University of Hong Kong

Background

Kernel density visualization (KDV) is a commonly used visualization tool for many data analytics tasks, including disease outbreak detection, crime hotspot detection, traffic accident hotspot detection, etc. Besides the data analytics tasks, IKCEST utilizes the heatmap, based on KDV, to show the distribution of COVID-19 cases in different regions of China to the general public. Although the most popular geographical information systems, e.g., QGIS, and ArcGIS, can also support this operation, these solutions are not scalable to generate a single KDV for datasets with million-scale data points, let alone to support interactive KDV (e.g., zoom in, zoom out, and panning operations) in near real-time (< 5 sec). To address this issue, we develop the interactive visualization tool IKDV-SYS that is built on top of our prior study [1] on the efficient kernel density computation. IKDV-SYS can achieve near real-time performance to generate KDV on large-scale datasets (e.g., New York traffic accident dataset with 1.3 million data points), under the single CPU setting.

IKDV-SYS

IKDV-SYS can be used to perform hotspot detection in different types of applications, e.g., traffic accident hotspot detection [2], crime hotspot detection [3,4], and disease outbreak detection [5,6]. In the following, we show the example usages of IKDV-SYS to detect hotspots (geographical regions with red/orange color) in different applications. Here, we also show the visualization results after we perform the zoom in operation. Based on this operation, IKDV-SYS can detect different hotspots in different levels (e.g., city level, street level, etc.).

Crime hotspot detection in Atlanta, using the open data from Atlanta police department


Traffic accident hotspot detection in New York, using the open data from New York government


COVID-19 hotspot detection in China, using the open data (February-March) from Johns Hopkins University


Compared with the state-of-the-art geographical information systems, e.g., QGIS, and ArcGIS, IKDV-SYS is the first web-based system, which can support:
(1) the near real-time kernel density visualization in single machine setting without using GPU and parallel computation
(2) different types of interactive visualization operations, e.g., zoom-in, zoom-out, panning etc, in near real-time
We have released our system (2nd version) to the general public. However, this is only the prototype in which we still need to refine it as better as possible. As such, we disclaim any liability in connection with the use of this system. In addition, since we cannot obtain the precise dataset for representing the COVID-19 cases, due to the privacy issue, the visualization in the city/street level is not accurate. As a remark, this system is a technical demo. The result presented in this system is not a reliable indicator of COVID-19 cases. Our demo paper is under submission to SIGMOD2021 (demo track) [7]. Hope that this paper can be accepted in this conference.

Video of IKDV-SYS

We have uploaded the video in Youtube for demonstrating our system.

Acknowledgement

Tsz Nam Chan, Shivansh Mittal and Reynold Cheng were supported by the Research Grants Council of HK (RGC Projects HKU 17229116, 106150091, and 17205115), the University of Hong Kong (Projects 104004572, 102009508, and 104004129), and the Innovation and Technology Commission of HK (ITF project MRP/029/18). Leong Hou U and Ye Li were funded by the National Key Research and Development Plan of China (No.2019YFB2102100), University of Macau (MYRG2019-00119-FST), the science and technology development fund, Macau SAR (SKL-IOTSC-2018-2020).

Contact Information

If you have any questions about our system, please feel free to contact Tsz Nam Chan, i.e., me (email: edisonchan2013928@gmail.com, just call me Edison) or (Ryan) Leong Hou U (email: ryanlhu@um.edu.mo).

References

[1] T. N. Chan, R. Cheng and M. L. Yiu. QUAD: Quadratic-bound-based kernel density visualization. In SIGMOD, pages 35-50. ACM, 2020.
[2] K. Xie, K. Ozbay, A. Kurkcu, and H. Yang. Analysis of traffic crashes involving pedestrians using big data: Investigation of contributing factors and identification of hotspots. Risk Analysis, 37(8):1459-1476, 2017.
[3] A. Ristea, M. A. Boni, B. Resch, M. S. Gerber, and M. Leitner. Spatial crime distribution and prediction for sporting events using social media. Int. J. Geogr. Inf. Sci., 34(9):1708-1739, 2020.
[4] T. Hart and P. Zandbergen. Kernel density estimation and hotspot mapping: examining the influence of interpolation method, grid cell size, and bandwidth on crime forecasting. Policing: An International Journal of Police Strategies and Management, 37:305-323, 2014.
[5] N. Muroga, Y. Hayama, T. Yamamoto, A. Kurogi, T. Tsuda, and T. Tsutsui. The 2010 foot-and-mouth disease epidemic in japan. The Journal of veterinary medical science / the Japanese Society of Veterinary Science, 74:399-404, 11 2011.
[6] P. Lai, C.-M. Wong, A. Hedley, S. Lo, P. Leung, J. Kong, and G. Leung. Understanding the spatial clustering of severe acute respiratory syndrome (sars) in Hong Kong. Environmental health perspectives, 112:1550-6, 12 2004.
[7] T. N. Chan, P. L. Ip, L. H. U, W. H. Tong, S. Mittal, Y. Li and R. Cheng. IKDV-SYS: A System for Near Real-Time Interactive Kernel Density Visualization. Under submission to SIGMOD, 2021 (demo track).