We were given a challenge by the sponsor McKinsey that consisted on a dataset of various sales from which we had to develop a forecasting model.

What it does

On our project we read, clean and reorganize the data, design the architecture of our model, deploy on Kaggle, train with sales data and finally forecast future sales.

How we built it

Python. Lybraries used are pandas, numpy and pytorch. Data handling is done locally, model deployment and training in Kaggle.

Challenges we ran into

Missing data values (a lot of NaNs), designing the architecture of our models and deployment.

Accomplishments that we're proud of

We have develop our own Deep Learning transformer model architecture.

What we learned

Making a forecasting model is hard.

Share this project: