TL;DR: I have a stepwise regression with adjusted R-square 0.9843 and I call it with an AWS API.
For this assignment, I am required to use SAS Enterprise Miner to create a regression model to predict the e-commerce sales of Korean women fashion products. This is a continuation from my previous assignment, which used the same data for data visualization. My data partition was set to 50% training and 50% validation, at 199 observations each. I achieved the highest adjusted R-square in my stepwise regression model. I used the Korean woman fashion Shopee Philippines dataset .
I joined the products and shops dataset. My stepwise regression model ended up with only 4 indepedent variables.
Since SAS Enterprise Miner is a propriety software, I could not find a way to directly deploy the model here, but I can use the intercept and estimates anyway, as in y=mx+c. In my old website, I put the formula behind a Django form, which is a bit clunky, since it reloads the entire page.
However, the intercept was less than -200. When all independent variables are at 0, a sales value of more than -200 simply does not make sense. So, I set it the prediction value to be 0 by default, if the prediction returns a negative value.