Project progress

It's been a busy week between meetings.  Olivia and Lily worked on our machine learning model and made some great progress.  If you remember, at the time of the regionals, we could predict showers with 90% accuracy.  Olivia and Lily improved that to 76% over the weekend.  Why is 76% an improvement over 90%, you ask?  Because although the new model heats water more times than necessary, it heats it almost every time it's supposed to.  So, as you can see in the chart below, out of 100 times someone comes into the bathroom, the system only fails to heat the water once.  It heats it 18 times when it doesn't need to, but existing systems heat it 100 times when they don't need to.

In addition, the new system runs on Amazon's computers.  That means that a computer with next to no intelligence can tell it that someone just came into the bathroom, and it will reply telling it whether to heat the water.  Cool!

In addition, Nic was hard at work designing circuit boards.  Here's a picture of a board we sent off to get manufactured for $1 each:

It would fit on top of an Arduino and replace the rats nest of wires we currently see.  The company we sent it to to get made said there are a couple of wires that cross each other that need to go around each other instead, and we need to change that.  That's not a big deal and we can probably fix it this week.  Good work Nic!

Olivia emailed one of the judges from the regionals who said she'd be happy to give us more feedback on our project presentation as we get ready for the Provincials.  That was very nice of her and could be a big help.

I also met with Gary Bennett, one of the property managers we were hoping to talk to before the regionals.  Some of you may remember him from Nature's Fury.  He was the mayor of Kingston during the Ice Storm and gave us some great ideas that year.  He said he'd be happy to help us again and hear more about our water saving innovation.