Im not sure if this is the correct category but here it goes.
I have 552 columns, each with a monthly figure of 552 different products. Id like to filter these and select only the ones that would be more suited for forecasting and exploring.
Im assuming that products with too many months of 0 sales would be the first filter as it means the product isnt sold too frequently.
A second filter would have to rely on those that have regular monthly sales but exhibit too much deviation from the mean although im not sure that is statistically correct. A product that exhibits a lot of growth in sales might deviate from the mean as the mean keeps moving.
A third filter might be how it correlates to other series or how it autocorrelates to itself, right?