Hackathon: Machine Learning with AWS
In spring 2019 I attended a three-day hackathon at the AWS offices in London. The purpose of the time was to use AWS Sagemaker to build a forecasting model to work out the best number of hotel rooms to contract in various countries across Europe. From this, we could apply logic to "swap beds" between markets to optimise supply to meet demand, and therefore maximise margin.
We experimented with various model types/approaches, including Linear Learner and XGBoost.
My team couldn't quite win the competition, ultimately finishing 2nd, but we learnt a lot about the tools that AWS offers, including getting hands on with Sagemaker, and playing with comprehend (natural language processing service) to see what it can do (with call centre transcripts).
It was great to dedicate time as a team to innovation and learning. The experimental work we did during the Hackathon has since been developed further to build a full "bed swaps" model.
See our presentation here.
