Looking to set-up a machine learning solution on AWS to predict outcomes in a dress-up game we operate.
We have a mobile dress-up game where users need to dress up their avatar for a particular event (e.g. beach party, christmas, dinner, visit to the mall).
Players choose dresses / shoes / makeup / etc from a set of options we present them. The player's composition is then submitted for voting.
Voting is done by other players in the game. Voting can last for upto a day.
The composition with the largest number of votes for that event, is the winner for that event.
We want to analyze which compositions + event pair generate the top votes.
Having 'learnt' these patterns, we then want an API that can be used to determine a composition's expected score, without having to wait for Voting to complete.
Aside from configuring the relevant learning models, training the system, building an end-point to predict outcomes, will also include scripts / process to collect the labelled data (that is already collected in a Postgres database outside of AWS).