search

facebook / prophet

Tool for producing high quality forecasts for time series data that has multiple seasonality with linear or non-linear growth.
https://facebook.github.io/prophet
MIT License
18.07k stars 4.49k forks source link

Is prophet even good for this type of tasks? #2476

Open mlinmg opened 11 months ago

mlinmg commented 11 months ago

I'm working on a forecasting problem where my client has data for around 5 years, but the sales events occur sporadically. These sales events typically last about a week out of a month, and both the dates and lengths of the events can vary widely.

I've been trying to use Prophet to forecast future earnings, but I'm encountering difficulties in capturing the right trends. I attempted to use the holidays functionality in Prophet to model these events, but since they can vary significantly, I'm concerned that Prophet might not be able to make accurate forecasts, maybe a better suit would be a DT, what do you think?

AlexandroLuis commented 11 months ago

Are you trying to use seasonality? it worked for me to get these gaps of high sales and low, here's an example: your_model.add_seasonality(name='weekly', period=30.5, fourier_order=6) # Adjusted Fourier order

mlinmg commented 11 months ago

Yes but this problem does not have seasonality(maybe like weekly but it's not standardised