Recently I started to create a machine learning model for a European customer for around 800 product time series. The goal is to produce a monthly forecast for the 6 months ahead.
Since this customer is a grocery wholesaler, a lot of the products experience supply chain difficulties due to Covid restrictions and now there might be some large effects due to the Ukraine crisis.
From the picture attached, you can already spectate the downwards trend in the last 6 months, which is pretty exceptional in most cases.
I did a lot of feature engineering and ran a bunch of algorithms (XGboost, Random Forest, SVM, Prophet with XGboost errors, Prophet with regressor).
In most cases the models fail to predict the downwards trend, which is actually not a big surprise.
I am searching for a more conservative approach, to achieve better accuracy and deal with uncertainty. One of my crude attempts was to multiply the forecast results by 0.5, which resulted in an accuracy almost twice as good.
If anyone has some ideas how to handle grocery wholesale forecasting during this challenging times, I would be glad for some advice.
Here some example time series:
