AppRecommender
AppRecommender copied to clipboard
tassia
A recommender system for GNU/Linux applications
Proposed by Joey Hess http://www.mail-archive.com/[email protected]/msg260184.html
Try different user profile strategies, such as maximal pakages list and subtracting basic packages (as stop list), and observe recommendation performance with them.
Create stop list for packages, based on common standard systems.
Consider information about package real utilization (last time is was used) when composing user's profile. The idea is: the less the package is used, the less it should weight on...
Consider number of open bugs, specially RC bugs, as a weight for final recommendation calculus.
Consider packages popularity provided by popcon reports as a weight for final recommendation calculus.
The recommender should consider the system impact of installing a package as a mean of weighting packages for recommendation. The less the impact, the higher the weight. Impact could be...