Pycspade

What is this?

This is a python wrapper for the C++ implementation of C-SPADE algorithm by the author, Mohammed J. Zaki Original code was downloaded from http://www.cs.rpi.edu/~zaki/www-new/pmwiki.php/Software/Software#toc11 Since this is just a wrapper it is as fast as the C++ code

How to install?

Compatible with Python 2 and 3. On Windows, Visual Studio 2015 Build Tools is also required.

pip install Cython pycspade

How to use?

Your data needs to be in a particular format similar to the following:

1 1 3 8 37 42
1 2 4 4 11 37 42
2 1 2 10 73
2 2 1 72
2 3 3 4 24 77
...

The first number is the sequence index, the second is the event index, the third is the number of elements, followed by the element, space separated

Let's call this file data.txt. You will call cspade as following:

from pycspade.helpers import spade, print_result

# To get raw SPADE output
result = spade(filename='tests/zaki.txt', support=0.3, parse=False)
print(result['mined'])

1 -- 4 4 
2 -- 4 4 
4 -- 2 2 
6 -- 4 4 
4 -> 6 -- 2 2 
4 -> 2 -- 2 2 
2 -> 1 -- 2 2 
4 -> 1 -- 2 2 
6 -> 1 -- 2 2 
4 -> 6 -> 1 -- 2 2 
4 -> 2 -> 1 -- 2 2

print(result['logger'])

CONF 4 9 2.7 2.5
args.MINSUPPORT 2 4
MINMAX 1 4
1 SUPP 4
2 SUPP 4
4 SUPP 2
6 SUPP 4
numfreq 4 :   SUMSUP SUMDIFF = 0 0
EXTRARYSZ 2465792
OPENED /tmp/cspade-WWv9bQWBYdDyH85T.idx
OFF 9 38
Wrote Offt 
BOUNDS 1 5
WROTE INVERT 
Cleaned up successful: /tmp/cspade-WWv9bQWBYdDyH85T.tpose
Cleaned up successful: /tmp/cspade-WWv9bQWBYdDyH85T.idx
Cleaned up successful: /tmp/cspade-WWv9bQWBYdDyH85T.data
Cleaned up successful: /tmp/cspade-WWv9bQWBYdDyH85T.conf

print(result['summary'])

CONF 4 9 2.5 2.7 10 1 4 0.781025 4
TPOSE SEQ NOF2 /tmp/cspade-WWv9bQWBYdDyH85T.data 0.3 4 2 1 
F1stats = [ 4 0 0 ]
SPADE /tmp/cspade-WWv9bQWBYdDyH85T.tpose 0.3 2 7 0 0 0 0 0 -1 1 100 100 4 5 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

# To also get other sequence mining's measures, incl. lift, support, confidence:
result = spade(filename='tests/zaki.txt', support=0.3, parse=True)
# Pretty print result:
print_result(result)

   Occurs     Accum   Support    Confid      Lift          Sequence
        4        14 1.0000000       N/A       N/A               (1) 
        4         6 1.0000000       N/A       N/A               (2) 
        2         4 0.5000000 0.5000000 0.5000000          (2)->(1) 
        2         2 0.5000000       N/A       N/A               (4) 
        2         2 0.5000000 1.0000000 1.0000000          (4)->(1) 
        2         2 0.5000000 1.0000000 1.0000000          (4)->(2) 
        2         2 0.5000000 1.0000000 1.0000000     (4)->(2)->(1) 
        2         2 0.5000000 1.0000000 1.0000000          (4)->(6) 
        2         2 0.5000000 1.0000000 1.0000000     (4)->(6)->(1) 
        4         6 1.0000000       N/A       N/A               (6) 
        2         4 0.5000000 0.5000000 0.5000000          (6)->(1) 

You can provide cspade with list of sequences instead of a file:

data = [
    [1, 10, [3, 4]],
    [1, 15, [1, 2, 3]],
    [1, 20, [1, 2, 6]],
    [1, 25, [1, 3, 4, 6]],
    [2, 15, [1, 2, 6]],
    [2, 20, [5]],
    [3, 10, [1, 2, 6]],
    [4, 10, [4, 7, 8]],
    [4, 20, [2, 6]],
    [4, 25, [1, 7, 8]]
]

result = spade(data=data, support=0.01)
print_result(result)

The result seq is a string, that have multiple rows and looks like this:

22 80 -> 72 -> 42 -> 22 -- 2 2
22 -> 45 71 -> 42 -- 1 1
80 -> 45 71 -> 42 -- 1 1
22 80 -> 45 71 -> 42 -- 1 1

Let's decipher the first row:

22 80 -> 72 -> 42 -> 22 -- 2 2

It gives you the frequent sequence followed by support (the last two numbers, which will be the same in this application). The row reads: the itemset (22 80) is followed by (72) followed by (42) followed by (22).

There're a lot of parameters that can be passed to this function. most important ones are:

• support: this is the minimum support level, default to 0 (not excluding anything)
• max_gap: The max number of itemset that can be skipped in a sequence
• min_gap: The min number of itemset that must be skipped in a sequence

Read the original paper and the C++ implementation for more details

How to contribute?

• Fork this repo
• Make change
• Pull request

How to recompile to use in IDE?

• rm cspade.cpp; python setup.py build_ext --inplace

Licence

• MIT