Artur Oliveira Rodrigues ; artur [at] dcc.ufmg.br
The documentation (in Portuguese) is located in the doc directory, and the
reference file is doc/tp1.pdf.
The algorithm was implemented in Python and its code can be found at
apriori.py.
The way the Apriori algorithm was implemeted allows the tuning of multiple parameters, as follows:
positional arguments:
-i, --input transactions file
-s, --support minimum support value [0, 1]
-c, --confidence minimum confidence value [0, 1]
optional arguments:
-a, --atm minimum value for the atm metric [0, 1]
-h, --help show this help message and exit
Here is a sample parameters configuration:
$ python apriori.py -i data/census_mod.dat -s 0.5 -c 0.7 -a 0.3
and here is the sample output (simplified for clarity):
Itemsets of size 1
country=United-States 0.897
(...)
workclass=Without-pay 0.000
Itemsets of size 2
country=United-States,race=White 0.788
(...)
edu_num=1,education=Preschool 0.002
Itemsets of size 3
country=United-States,race=White,salary<=50K 0.584
(...)
marital=Married-civ-spouse,race=White,relationship=Husband 0.365
RULES
edu_num=1 -> education=Preschool 1.000
(...)
country=United-States,workclass=Private -> race=White 0.880
(...)
age=middle-aged -> sex=Male 0.702