I’m new to Time Series and I have been working on Sales Forecasting problem, data is at week level from 2020 to 2024 – May, which contains normal sales and Promotion sales ( for a specific time period). Normal sales have a seasonality and model is able to capture it. But Promotional sales are different for each year and there is no proper seasonality present, model is not able to capture if there is sudden spike or very low sales present.
I have been experimenting with different FB prophet parameters but not much improvements.

I Have a problem with forecasting task, I want use facebook prophet, but I saw that doc don’t use train test split… Does Facebook prophet perform train test split automatically?