Development with what we have learned so far

In the first part of this project we explored different approaches to model sequences of metrics regarding S02 emissions. After basic EDA and simple modeling we will be building a smarter model.

Although convolutional network seem to give better results at first, after some experimenting I discovered it is not as good as recurrent stateful model. So here is what we will be bulding. But first we need to replicate the setup we have got before.

Setup repication