This README file describes the usage of data files and Mathematica modules in this directory, which accompanies the work "A method for forecasting the number of hospitalized and deaths based on the number of new infections during a pandemic", by R.Scitovski, K.Sabo, and Š.Ungar
Data sets Croatia.txt, Czechia.txt, and Hungary.txt contain data on new infected cases, deaths and hospitalizations in respective countries during the third wave of the COVID-19 pandemic (February to May 2021).
The form of each datum is {day, month, ti, ni, di, hi}, where:
day = day
month = month
ti = serial number
ni = the number of new infections on day ti
di = the number of new deaths on day ti
hi = the total number of hospitalizated on day ti
===========================================================================
Dead.nb
To use the supplied Mathematica notebook Dead.nb, first, under "Data input", as the argument of ReadList[" "] at the beginning of Dead.nb, enter either Croatia.txt, or Czechia.txt, or Hungary.txt (or the name of your own data-file).
Depending on the wanted output, there are three ways of using this notebook: "With details", "Without details", and "All training data sets":
"With details" will output:
- the bar charts of new infections and of deaths;
- the Logistic and Gompertz model functions;
- visualization of 90%, 95%, and 99% conﬁdence bands for the ﬁt of the function ϕ;
- the proportion of deaths among cases confirmed earlier;
- the forecast of the number of deaths during tau days starting on the day "T"
(the results are given separately for the Logistic and Gompertz model functions).
"Without details" will output only:
- Proportion of deaths among cases confirmed earlier;
- the forecast of the number of deaths during tau days starting on the day "T"
(the results are given separately for the Logistic and Gompertz model functions).
"All training data sets" will output the same as "Without details" for each "T" from "jIni" until "jEnd", plus the graphs of:
- deaths according to the Logistic model function from jIni until jEnd;
- deaths according to the Gompertz model function from jIni until jEnd;
- errors for Logistic and Gompertz model functions from jIni until jEnd;
- delays for Logistic and Gompertz model functions from jIni until jEnd;
- proportions of deaths for Logistic and Gompertz model functions from jIni until jEnd.
To use either "With details" or "Without details", one has to set "T" to the serial number "ti" of the starting date to output the number of predicted deaths (in our example: T=55). It is reasonable to set T>20 since the first 15 data are assumed to be related to deaths belonging to the previous pandemic wave. It may happen that for "T" around 20 the iterative process won't converge due to insufficient data.
The choice "With details" or "Without details" is done by setting either "Dead[DataT, 1]" or "Dead[DataT, 0]", respectively.
To use "All training data sets" one has to set "jIni" and "jEnd" to the serial numbers "ti" of the starting and of the final dates for which prognoses are requested.
===========================================================================
Hospitalized.nb
As with the Dead.nb, depending on the wanted output, there are three ways of using this notebook: "With details", "Without details", and "All training data sets":
"With details" will output:
- the bar charts of new infections and hospitalizations;
- the Logistic model function;
- visualization of 90%, 95%, and 99% conﬁdence bands for the ﬁt of the function ψ;
- the forecast of the number of hospitalizations during delta days starting on the day "T".
"Without details" will output only:
- the forecast of the number of hospitalization during delta days starting on the day "T".
"All training data sets" will output the same as "Without details" for each "T" from "jIni" until "jEnd", plus the graphs of:
- RMSE (root-mean-square error) from jIni to jEnd;
- MAPE (mean absolute percentage error) from jIni to jEnd;
- errors for the Logistic and Gompertz model functions from jIni to jEnd;
- delay from jIni to jEnd.
To use either "With details" or "Without details", one has to set "T" to the serial number "ti" of the starting date to output the number of predicted hospitalizations (in our example: T=65 for "With details" and T=20 for "Without details"). It is reasonable to set T>20 since the first 15 data are assumed to be related to deaths belonging to the previous pandemic wave. It may happen that for "T" around 20 the iterative process won't converge due to insufficient data.
The choice "With details" or "Without details" is done by setting either "Hospitalized[DataT, 1]" or "Hospitalized[DataT, 0]", respectively.
To use "All training data sets" one has to set "jIni" and "jEnd" to the serial numbers "ti" of the starting and of the final dates for which prognoses are requested.