DESCRIPTION

A recommender system for Debian packages. To make a recommendation, this application gathers the packages installed through apt and create an user profile based on them. This profile is then used to search for useful packages for a user.

This package provides several content-based recommendations approach and each one should provide different recommendations from the others.

GENERAL COMMANDS

-h, --help

Print this help

-i, --init

Initialize AppRecommender data by indexing the searchable packages in a Xapian database

-t, --train

Make training of AppRecommender machine learning strategies.

-s, --strategy=,OPTION/

Set recommendation strategy. If this flag is not used, the default strategy cb will be used as the default strategy value.

RECOMMENDATION STRATEGY OPTIONS

cb = content-based, mixed profile

cbt = content-based, tags only profile

cbd = content-based, description terms only profile

cbh = content-based, half-half profile

cbtm = content-based, time-context profile

cb_eset = cb with eset profiling

cbt_eset = cbt with eset profiling

cbd_eset = cbd_eset with eset profiling

cbh_eset = cbh with eset profiling

mlbva = machine_learning, Binary Vector Approach

mlbva_eset = machine_learning, Binary Vector Approach with eset profiling

mlbow = machine_learning, Bag Of Words

mlbow = machine_learning, Bag Of Words with eset profiling