Embedded Files
ABstract
ABstract
As a semester long project for DCS 3350 ("Contagion" - https://www.bowdoin.edu/~mirfan/DCS-3350.html) we attempt to model the spread of the novel Coronavirus (COVID-19) through mainland China using the Python package Networkx. Through these modeling attempts, we hope to gain some insight into how the virus has truly spread in China. Our model is far from perfect but some important conclusions can be drawn. Especially relevant in the spring of 2020, we implement our model to account for social distancing, and observe the impact from social distancing compared to the unregulated propagation of the disease. In both cases, our model produces numbers significantly greater than China's purported number of infected and number of deceased.
We expand on this topic by looking into the disease's social impact on China. Specifically, we look at the cultural and economic significance of Chinese wet markets, China's broader economy, and skepticism revolving around the Chinese government.
Authors: Jouya Mahmoudi '20 (jmahmoud@bowdoin.edu), Evan Haines '20 (ehhaines@bowdoin.edu), Ahmed Hameed '22 (ahameed@bowdoin.edu), Orlando Coyoy Ixquiacche '21 (ocoyoyix@bowdoin.edu)
Background
Background
In an effort to realize the immediate impact the novel Coronavirus has had on the world and the potential impact it may have on countries such as the US in the coming months, we attempt to model the spread of the virus through its assumed place of origin: China.
Like many other zoonotic diseases of recent, this strain of coronavirus (SARS-Cov-2) is thought to have developed among bats, and was passed to humans through an intermediary host animal. While the identity of this host animal remains largely uncertain, many speculate that pangolins - just one of many species of exotic animals sold at the Huanan Seafood Wholesale Market - are the culprit. Limited enclosure space at the market speculatively propagated the disease among the hosts, who transmitted the disease to humans upon consumption.
The pangolin is suspected to be the intermediary host for the coronavirus between bats and humans. Image from: https://www.afrik21.africa/en/africa-covid-19-gives-pangolin-a-break-from-extinction/
By the time of the market’s inevitable shutdown, the spread of the disease proved too vast to control. The first cases of the novel Coronavirus presented in December of 2019 in Wuhan. And through government censorship and misinformation, the disease’s scope increased mostly unregulated. After weeks had passed, the Chinese government finally acknowledged the existence of the disease, consequently quarantining various cities in the Hubei province.
Months later, China reports very few, if any, new infections of Covid-19 among its citizens, seeming to have capped its number of infected at slightly above 80,000. This, of course, has raised eyebrows - how is it that countries with populations only a portion of China’s, such as the USA, Spain, Italy, and France among others, have far exceeded China’s infected count?
It is our intention to also investigate this question, using our predictive model to approximate a more realistic number of infected Chinese citizens.
Our Model
Our Model
Our model focuses solely on the spread of the coronavirus through mainland China and is, for the lack of a better term, three layers “deep” - it models the spread of the virus at three scales: (1) Between provinces, (2) amongst different communities in a province, and (3) amongst sub-communities in a community.
First looking at the largest scope, the spread of the disease between provinces is dependent solely on air traffic data gathered from openflights.org. Edges are formed between provinces according to the passenger volume traveling between two respective provinces.
An example Barabasi-Albert Graph. Photo:https://barabasi.com/publications
Next, at the provincial level, nodes - representing different communities within the province - are connected by edged using the Barabasi-Albert Model. This is a preferential attachment network such that nodes with higher degrees more probabilistically form edges with newly born nodes.
With populations of each province in the millions, this is perhaps the most computationally expensive part of our model.
At the smallest scale, networks of sub-communities are formed using the Watts-Strogatz Model. This is appropriate, as members in a sub-community often display a high degree of connectivity. Graphically, these networks should display a "small-world" phenomenon.
Examples of the Watts-Strogatz Model. Photo: https://www.researchgate.net/figure/Algorithm-of-Watts-Strogatz-model-which-can-be-tuned-by-parameter-p-0-1_fig1_258120851
Page updated
Google Sites
Report abuse