Business AI - Sales predictor

Inspiration

We were very interested in Deeplearning and MachineLearning, so this was a very good oportunity for us to learn as much as possible.

What it does

It makes an estimation on what the sales would be for a product at a given situation. There's also an implementation to be used easily on single requests through a GUI.

How we built it

We made a Deep Learning model trained with a dataset made out most of the information given in the original datasets.

Challenges we ran into

For the front-end developers it was very challenging the fact that we had to make the interface using Python, a language that we hadn't used before.

In reference to the dataset itself, there was a lot of fiddling required in order to get something to work with which, in the end, also reflects on the performance of the model.

Aside from that, understanding the data was a challenge of itself, since it seemed as if every ID were referencing a sale (which lead to a whole night of making an unuseful estimation to fill in the apparently missing values), but ended up holding information in itself.

Accomplishments that we're proud of

Getting an aproach to Python, doing big progress in a very short time, designing insightful methods to fill in the blanks within the dataset and finding a way to merge all datasets into one through data referencing.

What we learned

Python programming, data processing, AI (DL) development (also discovered new ML approaches which helped us find the origin of some of the issues we were having).