0

I am analyzing sales transaction data for forecasting. Public Holidays and Weekend transaction values were '0' (Numeric Zero). To make my forecasts more accurate, how should I treat the public holidays/Weekends in the data? If I Omit Holidays data, am not getting proper model. Am using python for time series. Here is the Sample data

17-Dec-2018  3000
18-Dec-2018  3100
19-Dec-2018  3030
20-Dec-2018  3200
21-Dec-2018  3190 
22-Dec-2018     0
23-Dec-2018     0
24-Dec-2018  3020
25-Dec-2018     0
26-Dec-2018  2900 
27-Dec-2018  3030 
28-Dec-2018  3089
29-Dec-2018     0
30-Dec-2018     0
kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
Gower
  • 1
  • 1
  • there is no solution to detect lead, contemporaneous and lag effects of holidays/events in PYTHON . Please change your question or create a new one and also supply actual time series as https://stats.stackexchange.com/questions/313810/simple-method-of-forecasting-number-of-guests-given-current-and-historical-data/313852#313852 – IrishStat May 08 '19 at 18:08
  • If the data is seasonal, I've had success with GAM models, since weekends should be picked up by the day of week seasonality and holidays can be indicated by an additional variable. Here is a nice blog post walks through some examples [link](https://petolau.github.io/Analyzing-double-seasonal-time-series-with-GAM-in-R/) – Stephen G May 09 '19 at 02:46
  • Some similar posts (not duplicates): https://stats.stackexchange.com/questions/317192/forecasting-daily-time-series-with-many-zeros/317261, https://stats.stackexchange.com/questions/372125/forecasting-daily-time-series-sales-revenue-with-many-zero-entries. Or if using arima models in R, you can just recode the zeros as `NA`'s, R's arima can cope with that. For that see https://stats.stackexchange.com/questions/22910/how-to-deal-with-gaps-nans-in-time-series-data-when-using-matlab-for-autocorrela – kjetil b halvorsen May 10 '19 at 10:54

0 Answers0