Machine-Learning Customised Confectionary Selection Box

Everyone loves selection boxes, you shake the box, reach in and pull out a sweetie. Actually no, you look inside and rummage until you find the green one which is your favourite. Oh no! There are no green ones. Well the blue one will do... but none of those either! How come there's only purple and red ones left? They are horrid!

What happens is that every piece of packaging has a little RFID tag, and so the box monitors in real time how many of each colour are in it. It is a simple statistical calculation to give each colour a graduated preference ranking. This is stored and updated against your account at the wholesaler / retailer / manufacturer system.

Next time you go to the shop to purchase sweeties, as you stand at the sweety shelf looking at the different brands of boxes, you don't realise that all the boxes on display are empty fakes. As you reach out to pick up a box of PocBox sweety selection the iris recognition camera downloads your latest preference profile and quickly dumps the exactly appropriate proportion of each colour into the box at the front, and snaps the lid on just in time for you to lift it. You take it to the till and pay as usual etc. If you change your mind and reshelve the box the system discretely pulls it to the back of the shelf where it can be emptied back into the sorted storage hoppers.

As you guzzle through this custom box the system still watches the proportions in the box to gradually refine your selection and build seasonal variations (e.g. if you always buy a box in July for Aunt Constorphinol's next-door-neighbour-but-one the system will apply a new filter for that time of year and will not get confused by the very different guzzle pattern of Christmas when your stash of boxes is raided by hundreds of second and third cousins offspring