I could need a pointer, I have the feeling I am one inch away from understanding - but still something is missing. could you provide a hint or a resource or an sample implementation?

it's about creating a dynamic user model for my CMS. I observe the user and record it's interactions with the web page. I also managed to lift those interactions in a business domain - ie. interpret interactions with certain HTML Elements as interaction with a product / article / topic.

I got the data - as stream of filtered interactions - but still the math needed to create an actionable model eludes me. :(

I am looking for ways to aggregate the data into some expressions like "user got strong interest in blue lamps" the CMS could use to make decisions about content placement.

how can I organize the interaction in a way to draw sutch conclusion? what kinds of algorithm should I look into?

thx in advance, Merkur