- #1

- 92

- 0

I've been working on trying to model sales for my work and I'm really just modeling it using exponential regression, so I get the linear regression of the logarithm of the data and obtain the desired formula.

What I'm confused about is that if I integrate this to try to predict how many sales will be from now to some future point, it seems to work OK for products which have more sales (sale numbers are far from 1 for a given period of time) but not so well for products with poor sales (sales numbers that are either 0 or ~1 for a given period of time).

For the latter products the exponential model seems to predict that there are on average around 1 sale per unit of time, which when integrated ends up adding together a bunch of 1's for each period of time.

Am I misusing exponential regression or is this a limitation of its use?

Many thanks!