I am working on a sales dataset which has historical sales on a daily basis for 10 products (SKU) ranging through 2 years on a daily basis up to date (Including zeros for days with no sales). To predict the future sales (Say 30 days = 1 Month), I use prophet model by using a changepoint of .8 and adding the holidays for the country in which the data was obtained. The model works well but I am not able to capture the demand pattern well as this is the seven month using the predictions from prophet. To overcome this, I have thought of using a feedback loop as comparing the sales predicted vs the actual sales for the given period and use it to improve future predictions. How can I go about this?
$\begingroup$
$\endgroup$
Add a comment
|