---

MAFIA Code Documentation

Contact

• Manuel Calimlim
• Johannes Gehrke

Download

• Main MAFIA webpage http://himalaya-tools.sourceforge.net/Mafia

Linux Compilation

1. ./configure
2. make
   • executable is created as 'src/mafia'

Windows Compilation

1. Use cygwin and follow the Linux instructions or
2. Use a Windows Compiler or IDE (such as Visual Studio) to compile the source code.

Directory Structure

admin/	Contains config files for compiling. Should not be altered.
src/	Contains all of source code for MAFIA
src/Transaction.h src/Transaction.cpp	Class for reading transactions from ASCII datasets
src/ItemsetOutput.h src/ItemsetOutput.cpp	Class for writing itemsets to file output
src/BaseBitmap.h src/BaseBitmap.cpp	Simple bitmap class for name bitmaps
src/Bitmap.h src/Bitmap.cpp	Main bitmap class for transaction bitmaps
src/Mafia.cpp	Main class file with most of the MAFIA code
src/Tables.h	Stores precomputed lookup tables (not included in documentation due to very large tables)
src/TreeNode.h	Class for representing nodes in the search tree
INSTALL	Generic installation instructions for Linux/Unix
mafia.{kdevprj,kdevses}	KDevelop project files for Linux
README	Pointer this page

Program Usage

    Usage: mafia [-mfi/-fci/-fi] [min sup (percent)]
            [-ascii/-binary] [input filename]
            [output filename (optional)]
    Ex: mafia -mfi .5 -ascii connect4.ascii mfi.txt
    Ex: mafia -mfi .3 -binary chess.binary
    

File Input

Datasets can be in ASCII or binary format. For ASCII files, the file format must be:

    [item_id_1] [item_id_2] ... [item_id_n]
    

Items do not have to be sorted within each transaction. Items are separated by spaces and each transaction should end with a newline, e.g.

    1 4 2
    2 8 9 4
    2 5
    

For binary files, the file format must be:

    [custid][transid][number of items][itemid_1][itemid_2] ... [itemid_n]
    

The custid and transid numbers are ignored at this time. Since the file is in binary format, all numbers are read as integers.

Datasets

Download Datasets-ascii.tar.gz or Datasets-binary.tar.gz for the full set of datasets used for testing:

• chess.{ascii,binary}
• connect4.{ascii,binary}
• mushroom.{ascii,binary}
• pumsb.{ascii,binary}
• pumsb_star.{ascii,binary}

Program Output

The frequent itemsets outputted by the program are in ASCII form with the following format:

    [item_id_1] [item_id_2] ... [item_id_n] [(support)]
    

Ex:

    28 64 42 60 40 29 52 58 (966)
    46 64 42 3 25 9 5 48 66 56 34 62 7 36 60 40 29 52 58 (962)
    39 36 40 29 52 58 (960)
    

---