Holidays ,weekends and seasonality pattern in time series

[ash25sumbre]

I have data which is based on daily sales . 

the process  which I followed:-

How I performed analysis I'm giving you approach

1. the data is from April 2019 till May 2021

2.I've created training and testing data using 80/20 rule

3. After that I've checked ETS AND ARIMA MODELS on training data.

4. so after running 2 models check for RMSE values and the model which is having lower RMSE value I'm using that for forecasting (eg. in ARIMA and ETS  if RMSE value is lower for ARIMA model I'm using Arima and run whole data using ARIMA and also forecast the values using it.)

Now I want to include holidays /weekend logic to data and then want to forecast how should I proceed?  Please help me with this.

[mceleavey · Accepted solution]

Hi @ash25sumbre ,

It looks like you need to filter out the weekends from your data.

I've attached a workflow which generates every date between those two dates and removes the weekends.

As for the holidays, it depends on which countries you are using, which holidays in particular you wish to remove and what dates they are.

Assuming you have this data, you simply need to join the holiday dates to the dates in the workflow and take the dates that don't match the holidays.

I hope this helps,

M.

[ash25sumbre]

Thank you for  your help . 

[mceleavey]

[anniephung]

@mceleavey thank you for the suggestion! i'm also following the same approach except now im not sure how to forecast for only weekend dates? if you have any pointers, i'd be grateful! thank you!