Sales Prediction using RNN by Gommet

Inspiration

We really liked data science, statistics, machine and deep learning applications and actually some of us work in these studies. Therefore, it could be a better idea to give a chance to that project and learn new models such as LSTM and RNN.

What it does

We give an ensembled sales predictive model which predicts the sales of a determined date given the category product, its price and its former sales.

How we built it

We built our predictive model by tensorflow, pandas and scikit even though we had spent some time with pytorch as it is the technology most used by us.

Challenges we ran into

One of the most challenging we ran into was considering the windows of the data, as it is a time series problem. Moreover, we had some troubles with pytorch and even though we tracked the learning curve by tensorboard, we started to use tensorflow.

Accomplishments that we're proud of

Prouds of having a good metrics such MSE, of being capable to understand the dataset and the knowledge gained by doing statistical and deep learning models.