A fun and very useful competition as well.

Congratulations to the winners and good luck to everyone.

here is my solution in simple details:

First, we group the data by date and source, then sum up the Kilowatt * Hr. of each day for each source, finally we perform the division into 5 days feature and label samples.

The model can be divided into two main parts:

- the first part is the parallel neural network models with the two different inputs entering to it

- the second part is the concatenation layer and then the Dense layer for getting output

Here you can find two charts that visualizes these ideas, thank you!

https://imgur.com/a/SYiMw0Z