0
$\begingroup$

I have 2 datasets from 2013-2017 for each day. a) Revenue generated by Locations and date. b) Holiday name and date

I would like to know how each holiday is impacting the revenue by location. I am confused about what method to use! Should 3-Sigma rule or Bollinger plot be enough or should I try spline regression or something else? FYI I am using python for this analysis.

$\endgroup$
2
  • $\begingroup$ Welcome to the community @Rohan. The solution to your problem has a wide scope and can cause the question to be closed. Please try to elaborate on a specific problem rather than the overall concept. $\endgroup$ – thanatoz Dec 2 '18 at 20:03
  • $\begingroup$ I am trying to forecast the revenue by location and day. I am trying to use Holidays as one of the predictor variable. So I am trying to give some weight to each Holiday depending upon how they have impacted the revenue in the past. $\endgroup$ – Rohan Singh Dhaka Dec 2 '18 at 20:08
0
$\begingroup$

It might be not the best approach, but without knowing the data its hard to be precise. If you have no ideas where to start, I can suggest trying this things first.

  1. If you want to get the "impact" of the holiday, try to one-hot encode "holiday" variable, and use some simple model that is interpretable. You need to undestand how each holiday-feature impacts the revenue, maybe linear regressiong is a good way to start and move on to more advanced algo`s.
  2. You can try to play with Facebook`s Prophet - its an package for timeseries predictions, wich shows the daily impacts by default. Here is the link: https://research.fb.com/prophet-forecasting-at-scale/ . And link to python package: https://github.com/facebook/prophet

Additional post about Propet and transactios forecasting: https://www.kaggle.com/armamut/predicting-transactions-fb-prophet-tutorial

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.