I am using R for analysis. My data is as follows:
id timestamp cumsum
1284381 21/01/2015 33
1284381 21/01/2015 57
1284381 2/3/2015 79
1284381 4/3/2015 203
1284381 25/03/2015 475
1284381 11/4/2015 578
1284381 17/04/2015 856
1284381 21/04/2015 1189
1284381 5/5/2015 1214
1284381 10/5/2015 1321
1284381 12/5/2015 1340
1284381 15/05/2015 1529
1284381 18/05/2015 1649
1284381 19/05/2015 1977
1284381 21/05/2015 2385
1284381 23/05/2015 2528
1284381 26/05/2015 2556
1284381 29/05/2015 2705
1284381 1/6/2015 2898
1284381 4/6/2015 2913
1284381 7/6/2015 2921
1284381 13/06/2015 2922
1284381 13/06/2015 3622
1284381 16/06/2015 3834
1284381 19/06/2015 3913
1284895 27/01/2015 6
1284895 27/01/2015 49
1284895 18/03/2015 57
1284895 20/03/2015 58
1284895 23/03/2015 59
1284895 23/03/2015 60
1284895 24/03/2015 62
1284895 29/03/2015 67
1284895 31/03/2015 75
1284895 1/4/2015 76
1284895 2/4/2015 77
1284895 8/4/2015 78
1284895 16/04/2015 80
1284895 21/04/2015 103
1284895 23/04/2015 275
1284895 26/04/2015 293
1284895 27/04/2015 386
1284895 30/04/2015 539
1284895 3/5/2015 807
1284895 8/5/2015 851
1284895 11/5/2015 988
1284895 14/05/2015 1056
1284895 18/05/2015 1157
1284895 21/05/2015 1226
1284895 23/05/2015 1383
1284895 26/05/2015 1501
1284895 30/05/2015 1518
1284895 2/6/2015 1694
1284895 4/6/2015 1695
1284895 8/6/2015 1858
1284895 11/6/2015 1909
1284895 14/06/2015 1917
1284895 17/06/2015 1957
1284895 20/06/2015 1973
The first column is ID, second is date and third is cumulative sum of the value. I want to build a forecasting model to this data, which can provide me a solution of, for a given id, at a future date(say. 08/08/2015), the cumsum would be ?? I have tried forecasting models with two variables. Since it is three variables and also the data is daily data and not continuous, I am facing difficulties in setting up the model.
This should be fairly straightforward, but I'm sure you'll want to tweak this for more detail. Look at the
forecast
package for more information. It's a great tool.Sample Data :
Code :
Output :