Arima models — Corona virus — Orange data mining
Today we will talk how to use arima time series models for predictions for corona virus. We will use Orange data mining tool.
1) Data downloading:
You can download data using the link: https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_confirmed_global.csv
2) Transpose our data using module transpose ( choose country/region)
3) Choose specific country for investigation with select columns widget — Russia for example.
4)We will build time series starting from the point when there are at least 10 observations. So we chose subset of data from 10 observation until the end. Here we use data table module.
5) Now we need to find the best parameters for Arima model. Here we will use python script module and code.
The code can be downloaded from:
6) We execute the script and inside the script we will get the parameters of the best model. The best ARIMA model for Russia for date up to day is Arima(0,2,1).
Now we can build our predictions:
7) We attach the widgets: select columns to data table, arima model to select column and data table to arima model:
In select models we move the Russia from feature to target.
In arima models we choose parameters of optimal model 0,2,1 and 10 for the number of forecasting
And we can see our prediction for the next 10 days in data table
That all