IEEE Computational Intelligence Magazine - November 2020 - 21

two parts: hospital and home. To isolate suspected infections in
hospitals, many hospitals have been utilizing robots to serve
patients that arrive. For example, at Antwerp University Hospital, Cruzr Health [61] takes patients' body temperatures and
checks whether they are wearing masks when patients arrive. If
the patients are wearing masks and their body temperatures are
within the normal range, Cruzr Health leads them to their
appointment. To take care of home isolation, Canada developed
a chatbot [62] that provides information for home isolation and
reminders for those who are suspected to be infected. As shown
in Fig. 3, we can see that the chatbot provides information on
COVID-19 symptoms at users' requests.
At the end of the pandemic, the most important task for
every country would be economic recovery. However, the timing of enforcing policies for economic recovery is a difficult
problem. Researchers at WeBank [64], a private Chinese neobank, collected data from satellites, GPS, and social media to
detect the hot spots of actual steel manufacturing inside the
factories in China [63]. They believe that the detection results
may reflect economic recovery in China. Although this system
was built for internal use, the data from satellites, GPS, and
social media can be crawled easily, implying that computational intelligence researchers can build a model according to
this concept.
IX. Conclusions

In this survey, we reviewed several critical issues on combating
COVID-19 that have been or can be resolved using computational intelligence techniques. Computational intelligence is
classified into five different principles: neural networks, fuzzy
logic, evolutionary computation, computational learning theory, and probabilistic methods. Our survey found that most
research studies have been designed based on neural networks
for addressing the issues on characterization of the symptoms
of viral infections. Meanwhile, Panwar et al. [16]'s method
achieved the highest performance (97.62% true-positive rate),
which means that using deep neural networks to detect symp-

toms from CT images is well-developed, and we may devote
our efforts to other issues.
Theoretically, all issues we listed in Section II can be solved
by at least one of the principles of computational intelligence.
Unfortunately, based on our survey, many COVID-19 pandemic
issues have not yet been addressed in computational intelligence
studies. On the contrary, most reported studies have focused only
on specific issues, such as the characterization of the symptoms of
viral infection. This may be because computational intelligence is
a data-driven technique that can work well mostly when the
amount of data is sufficient. Currently, the data that we can
most easily crawl is chest CT images. Therefore, existing works
have focused on discovering the characteristics of COVID-19
patients based on their chest CT images to build classification
models. As more and more patients are cured, many treatment
records will be produced. Such treatment records could be
viewed as a set of time series data. Many computational intelligence techniques could be then applied to analyze treatment
records. To address the issues on TrD and PD, future works can
combine time-series analysis mechanisms with previous works.
For example, if we obtain COVID-19 patients' CT images for
each stage, the characteristics at each stage can be modeled and
utilized for treatment design.
Finally, we observe that some existing works, such as [42],
[46]-[48], utilized more than two principles to design hybrid
models that can balance the strengths and weaknesses of two
principles so that the applicability of these works could be
improved. For example, evolutionary computation could be
used to optimize the hyperparameters of deep learning models
so that some deep learning models might be built from limited
data. We believe, in the near future, the computational intelligence community will invent new algorithms by combining
multiple principles to address the critical issues described in this
survey using limited data or under strict conditions, such as
visual analytics techniques and applications for propagation
modeling and monitoring, vaccine design or drug repositioning, as well as IoT for smart care in COVID-19.
Acknowledgment

This research was partially supported by Ministry of Science
and Technology Taiwan under grant no. MOST 109-2224-E009-003 and by T.T. and W.F. Chao Foundation and John S
Dunn Research Foundation at Houston, Texas, USA.
References

FIGURE 3 A snapshot of Canada's COVID-19 chatbot.

[1] F.S. Lu et al., "Accurate inf luenza monitoring and forecasting using novel internet data
streams: A case study in the Boston metropolis," JMIR Public Health Surveill., vol. 4, no. 1,
p. e4, 2018. doi: 10.2196/publichealth.8950.
[2] S. Jang, S. Lee, S.M. Choi, J. Seo, H. Choi, and T. Yoon, "Comparison between SARS
CoV and MERS CoV using Apriori algorithm, decision tree, SVM," in Proc. MATEC
Web Conf., 2016, vol. 49, p. 08001. doi: 10.1051/matecconf/20164908001.
[3] D. Kim, S. Hong, S. Choi and T. Yoon, "Analysis of transmission route of MERS
coronavirus using decision tree and Apriori algorithm," in Proc. 18th Int. Conf. Advanced
Communication Technology (ICACT), Pyeongchang, 2016, pp. 559-565.
[4] "COVID-19 Open Research Dataset Challenge (CORD-19)." Accessed: Aug. 3,
2020. [Online]. Available: https://www.kaggle.com/allen-institute-for-ai/CORD-19
-research-challenge
[5] J. C. Bezdek, "What is computational intelligence?" in Computational Intelligence: Imitating Life, J. Zurada, R. Marks, and C. Robinson, Eds. Piscataway, NJ: IEEE Press, 1994,
pp. 1-12.

NOVEMBER 2020 | IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE

https://www.kaggle.com/allen-institute-for-ai/CORD-19-research-challenge https://www.kaggle.com/allen-institute-for-ai/CORD-19-research-challenge

IEEE Computational Intelligence Magazine - November 2020

Table of Contents for the Digital Edition of IEEE Computational Intelligence Magazine - November 2020