We visualized the number of confirmed cases on a map. We start by loading the latitude and longitude variables.

opts = detectImportOptions(filenames(4), "TextType","string");
Copy the code

The dataset contains “province/state” variables, but we aggregate the data at the “country/territory” level. Before we do that, we need to sort out the data a little bit.

times_conf.("Country/Region")(times_conf.("Country/Region") == "China") = "Mainland China";
times_conf.("Country/Region")(times_conf.("Country/Region") == "Czechia") = "Czech Republic";
Copy the code

We can now use groupsummary to add up confirmed cases and average latitude and longitude to aggregate the data by country/region.

country = groupsummary(times_conf,"Country/Region",{'sum','mean'},vars(3:end));
Copy the code

The output contains unnecessary columns, such as the sum of latitude and longitude. Let’s delete these variables.

vars = regexprep(vars,"^(sum_)(? =L(a|o))","remove_"); vars = regexprep(vars,"^(mean_)(? =[0-9])","remove_");Copy the code

times_conf_exChina = times_conf_country(times_conf_country.("Country/Region") ~= "Mainland China",:);
Copy the code

Let’s visualize the first and last date data in the dataset using GeoBubble.

for ii = [4, length(vars)]
    times_conf_exChina.Category = categorical(repmat("<100",height(times_conf_exChina),1));
    times_conf_exChina.Category(table2array(times_conf_exChina(:,ii)) >= 100) = ">=100";
    gb.LegendVisible = "off";
Copy the code

We can see that it initially affected only the countries around the continent. It is important to note that we have confirmed cases in the United States as early as January 22, 2020.

Let’s compare the number of confirmed cases by country/region using COVID_19_data.csv. There are inconsistencies in the date-time format, so we treat it as text at first.

opts = detectImportOptions(filenames(3), "TextType","string","DatetimeType","text");
Copy the code

Clear date and time format.

Data.nDate = regexprep(Data.Date,"\/20$","/2020");
Data.Date = datetime(Data.Date);
Copy the code

We also need to standardize values in country/region.

Country_Region(Country_Region == "Iran (Islamic Republic of)") = "Iran";
Copy the code

The dataset contains provincial/state variables. Let’s aggregate the data at the country/region level.

countryData = groupsummary(provData,{'ObservationDate','Country_Region'}, ...
    "sum",{'Confirmed','Deaths','Recovered'});
Copy the code

CountryData contains cumulative daily data. We just need the latest numbers.

We can calculate the number of new cases by subtracting the cumulative number of confirmed cases between the two dates.

for ii = 1:length(labelsK)
    country = provData(provData.Country_Region == labelsK(ii),:);
    country = groupsummary(country,{'ObservationDate','Country_Region'}, ...

    if labelsK(ii) ~= "Others"
        nexttile
Copy the code

As you can see, China and South Korea are not seeing many new cases. We can see that the epidemic has been contained.

The number of valid cases is falling, and the curve looks roughly gaussian. Can we fit the Gaussian model and predict when the activity case will be zero?

I use the curve fitting toolbox for Gaussian fitting.

ft = fittype("gauss1");

[fobj, gof] = fit(x,y,ft,opts);
gof
Copy the code

Gof = struct with fields: sse: 4.4145e+08 rsquare: 0.9743 dfe: 47 adjrsquare: 0.9732 rmse: 3.0647e+03Copy the code

Let’s output the forecast by adding 20 days.

Now let’s plot the result.

figure
area(ObservationDate,by_country{1}.Active)
hold on
plot(xdates,yhat,"lineWidth",2)
Copy the code