Bnr is a command-line tool for bipartite network recommendation.

libbnr is the library bnr is using to do its job.

Algorithms included currently:

• mass
• heat
• hybrid
• HNBI
• mass degree
• mass score 1
• mass score 2

TODO

1. git clone https://github.com/RuiOAIX/BNR
2. cd BNR && make
3. The command bnr lives in bin folder.
4. The header of libbnr lives in include folder, and the static library libbnr.a lives in lib folder.

1. ./run -i ~/Downloads/movielens_3c.txt -S

• Common options:

  • -h: help
  • -o logfilename:
    • Setting the log file

• Options privated to Algorithm:

  • -m: Calculate the result of mass algorithm
  • -e: Calculate the result of heats algorithm
  • -H: Calculate the result of hybrid algorithm
  • --hybrid-rate doubleValue:
    • Rate used in hybrid algorithm.
    • only valid when -H option is used, otherwize this arg will be ignored.
    • default: 0.2
  • -N: Calculate the result of HNBI algorithm
  • --HNBI-rate doubleValue:
    • Rate used in HNBI algorithm.
    • only valid when -N option is used, otherwize this arg will be ignored.
    • default: -0.8
  • -D: Calculate the result of mass degree algorithm
  • --mass-degree-rate doubleValue:
    • Rate used in mass score algorithm.
    • only valid when -D option is used, otherwize this arg will be ignored.
    • default: -0.8
  • -S: Calculate the result of mass score algorithm
  • --mass-score-rate doubleValue:
    • Rate used in mass score algorithm.
    • only valid when -S option is used, otherwize this arg will be ignored.
    • default: -0.8
  • -w: Calculate the result of mass score (only the third step) algorithm
  • --mass-score-third-rate doubleValue:
    • Rate used in mass score (only the third step algorithm.
    • only valid when -w option is used, otherwize this arg will be ignored.
    • default: -0.8

• Options related to dataset & file:

  • -i filename:
    • File containing full dataset
    • if -i is used, then -T and -t will be ignored.
    • else, -T and -t both have to be present.
  • -T filename:
    • File containing train dataset
  • -t filename:
    • File containing test dataset
  • -u filename:
    • File containing extra attribute of the recommending objects
  • -d doubleValue:
    • Rate used to divide full dataset to train and test dataset
    • only valid when -i option is used

• Other options:

  • -l intValue:
    • Number of times which the algorthim calculation need to be performed
    • in order to get reasonable average result
  • -L intValue:
    • Number of the recommended objects which will be used to computer metrics
  • -s unsignedlongValue:
    • Random seed