-
-
a demo run
Inspiration
My home country does not have an optimized product ordering system. The orders come in and out, sometimes disappearing for months; The managers them selves don't have a nice attitude towards others as well.
What it does
As the title suggests, the model orders products for you based on the last 7 day's sales, the time limit you want it to be sold and what amount of products are currently at the store.
How we built it
I used pytorch and numpy. The data it self was generated and randomized between specific parameters and formulas; And it was fine tuned to be 91% successful at ordering the right amount of products.
Challenges we ran into
In calculating the core logic, i found out that it would be more beneficial to be under-stocked than overstocked. Since products that are overstocked gets returned of its not sold well and eventually be a waste of product that could have been sold in another store, i made the parameters between 0.5*(the time limit you want it to be sold) to (the time limit you want it to be sold).
Accomplishments that we're proud of
This is my 3rd successful AI project.
What we learned
That Layernormalizers, (reward - reward mean) / reward std and dropouts play a crucial role in a model.
What's next for Product manager
I will train it to include events as well. Like on Christmas the sales increase by 1.5.
Log in or sign up for Devpost to join the conversation.