I am trying to predict monthly sales by product based on a plethora of variables. There are 4 predictors. One is categorical (month) and the other three are numerical. One of the variables is just part sales.
The data I am trying to predict is nested — there are product groups and colors within each product group. I am trying to predict sales by color.
The data is cyclical (sales vary by month/season).
What model would you recommend using?
You can try out hierarchical time series forecasting.
Or treat each color wise product wise sales forecasting. There are many auto ml time series packages such as autoarima and pycaret.
I would break out each product variant (color, size, features) into its own category and use VAR (Vector Auto Regression) It is a nifty form of time series which takes into consideration the typical time series stuff, but also builds in regressions for each predictor (weighted by lag periods) that computes the influence of each of the predictors on the others and the outcome at future times.
It is a useful way of doing multivariate time series forecasting!
